chore(cd): retry post-deploy smoke

.claude/settings.json

@@ -1,832 +0,0 @@
-{
-  "permissions": {
-    "allow": [
-      "Read(**)",
-      "Glob(**)",
-      "Grep(**)",
-      "Bash(curl *)",
-      "Bash(kubectl get *)",
-      "Bash(kubectl describe *)",
-      "Bash(kubectl logs *)",
-      "Bash(kubectl rollout status *)",
-      "Bash(docker ps *)",
-      "Bash(docker logs *)",
-      "Bash(ls *)",
-      "Bash(cat *)",
-      "Bash(head *)",
-      "Bash(tail *)",
-      "Bash(grep *)",
-      "Bash(find *)",
-      "Bash(pwd)",
-      "Bash(which *)",
-      "Bash(echo *)",
-      "Bash(git status *)",
-      "Bash(git log *)",
-      "Bash(git diff *)",
-      "Bash(git branch *)",
-      "Bash(git remote *)",
-      "Edit(**)",
-      "Write(apps/**)",
-      "Write(packages/**)",
-      "Write(docs/**)",
-      "Write(.agents/**)",
-      "Write(k8s/**)",
-      "Write(scripts/**)",
-      "Bash(pnpm *)",
-      "Bash(npm *)",
-      "Bash(npx *)",
-      "Bash(node *)",
-      "Bash(python *)",
-      "Bash(python3 *)",
-      "Bash(pip *)",
-      "Bash(cd *)",
-      "Bash(mkdir *)",
-      "Bash(touch *)",
-      "Bash(cp *)",
-      "Bash(mv *)",
-      "Bash(chmod *)",
-      "Bash(pytest *)",
-      "Bash(playwright *)",
-      "Bash(git add *)",
-      "Bash(git commit *)",
-      "Bash(git stash *)",
-      "Bash(ssh *)",
-      "Bash(scp *)",
-      "Bash(export KUBECONFIG=*)",
-      "Bash(git push:*)",
-      "Bash(claude --version)",
-      "Bash(git check-ignore:*)",
-      "WebSearch",
-      "Bash(claude plugin:*)",
-      "Bash(claude --channels)",
-      "Bash(claude --channels plugin:telegram@claude-plugins-official --help)",
-      "Bash(bash)",
-      "Bash(source ~/.zshrc)",
-      "Bash(~/.bun/bin/bun --version)",
-      "Bash(env)",
-      "Bash(claude upgrade:*)",
-      "Bash(/Users/ogt/.local/bin/claude --help)",
-      "Bash(CLAUDE_CODE_EXPERIMENTAL_CHANNELS=1 claude --help)",
-      "Bash(claude --channels plugin:telegram@claude-plugins-official --print \"hello\")",
-      "Bash(mkdir -p ~/.claude/channels/telegram)",
-      "Bash(~/.claude/channels/telegram/.env)",
-      "Bash(~/.bun/bin/bun run:*)",
-      "Bash(sudo ln:*)",
-      "Bash(ln -sf ~/.bun/bin/bun /opt/homebrew/bin/bun)",
-      "Bash(xargs python:*)",
-      "Bash(uv --version)",
-      "Bash(pip3 install:*)",
-      "Bash(pip3 show:*)",
-      "Bash(ruff *)",
-      "Bash(mypy *)",
-      "Bash(black *)",
-      "Bash(isort *)",
-      "Bash(timeout *)",
-      "Bash(wc *)",
-      "Bash(sort *)",
-      "Bash(uniq *)",
-      "Bash(awk *)",
-      "Bash(sed *)",
-      "Bash(tr *)",
-      "Bash(tee *)",
-      "Bash(xargs *)",
-      "Bash(test *)",
-      "Bash([ *)",
-      "Bash(true)",
-      "Bash(false)",
-      "Bash(date *)",
-      "Bash(sleep *)",
-      "Bash(kill *)",
-      "Bash(pkill *)",
-      "Bash(ps *)",
-      "Bash(top *)",
-      "Bash(htop *)",
-      "Bash(df *)",
-      "Bash(du *)",
-      "Bash(free *)",
-      "Bash(uname *)",
-      "Bash(hostname *)",
-      "Bash(whoami)",
-      "Bash(id *)",
-      "Bash(groups *)",
-      "Bash(stat *)",
-      "Bash(file *)",
-      "Bash(realpath *)",
-      "Bash(dirname *)",
-      "Bash(basename *)",
-      "Bash(type *)",
-      "Bash(command *)",
-      "Bash(hash *)",
-      "Bash(alias *)",
-      "Bash(set *)",
-      "Bash(unset *)",
-      "Bash(printenv *)",
-      "Bash(diff *)",
-      "Bash(cmp *)",
-      "Bash(comm *)",
-      "Bash(join *)",
-      "Bash(paste *)",
-      "Bash(cut *)",
-      "Bash(rev *)",
-      "Bash(nl *)",
-      "Bash(fmt *)",
-      "Bash(fold *)",
-      "Bash(pr *)",
-      "Bash(expand *)",
-      "Bash(unexpand *)",
-      "Bash(od *)",
-      "Bash(xxd *)",
-      "Bash(hexdump *)",
-      "Bash(strings *)",
-      "Bash(base64 *)",
-      "Bash(md5sum *)",
-      "Bash(sha256sum *)",
-      "Bash(jq *)",
-      "Bash(yq *)",
-      "Bash(gh *)",
-      "Bash(docker build *)",
-      "Bash(docker run *)",
-      "Bash(docker exec *)",
-      "Bash(docker compose *)",
-      "Bash(docker-compose *)",
-      "Bash(docker images *)",
-      "Bash(docker inspect *)",
-      "Bash(docker network *)",
-      "Bash(docker volume *)",
-      "Bash(kubectl apply *)",
-      "Bash(kubectl create *)",
-      "Bash(kubectl exec *)",
-      "Bash(kubectl port-forward *)",
-      "Bash(kubectl config *)",
-      "Bash(helm *)",
-      "Bash(terraform *)",
-      "Bash(ansible *)",
-      "Bash(bun *)",
-      "Bash(deno *)",
-      "Bash(cargo *)",
-      "Bash(rustc *)",
-      "Bash(go *)",
-      "Bash(java *)",
-      "Bash(javac *)",
-      "Bash(gradle *)",
-      "Bash(mvn *)",
-      "Bash(make *)",
-      "Bash(cmake *)",
-      "Bash(ninja *)",
-      "Bash(uv *)",
-      "Bash(poetry *)",
-      "Bash(pipx *)",
-      "Bash(virtualenv *)",
-      "Bash(venv *)",
-      "Bash(conda *)",
-      "Bash(brew *)",
-      "Bash(apt *)",
-      "Bash(apt-get *)",
-      "Bash(yum *)",
-      "Bash(dnf *)",
-      "Bash(pacman *)",
-      "Bash(snap *)",
-      "Bash(flatpak *)",
-      "Bash(systemctl status *)",
-      "Bash(journalctl *)",
-      "Bash(service * status)",
-      "Bash(nc *)",
-      "Bash(netstat *)",
-      "Bash(ss *)",
-      "Bash(lsof *)",
-      "Bash(nmap *)",
-      "Bash(dig *)",
-      "Bash(nslookup *)",
-      "Bash(host *)",
-      "Bash(ping *)",
-      "Bash(traceroute *)",
-      "Bash(mtr *)",
-      "Bash(wget *)",
-      "Bash(http *)",
-      "Bash(httpie *)",
-      "Bash(hadolint apps/api/Dockerfile)",
-      "Bash(docker info:*)",
-      "Bash(kubectl cluster-info:*)",
-      "Read(//var/run/**)",
-      "Bash(open -a Docker)",
-      "Bash(git rm:*)",
-      "Bash(git reset:*)",
-      "Bash(kubectl --kubeconfig ~/.kube/config get pods -n awoooi -o wide)",
-      "Bash(kubectl scale:*)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollam@192.168.0.188 \"docker ps -a | grep -i claw\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps -a | grep -i claw\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker start clawbot && sleep 3 && docker logs clawbot --tail=10\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps | grep clawbot && docker port clawbot\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot --tail=30\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cat /home/ollama/clawbot/.env | grep -E ''\\(TG_|TELEGRAM\\)'' | head -5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker inspect clawbot --format=''{{range .Mounts}}{{.Source}}:{{.Destination}} {{end}}''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker inspect clawbot --format=''{{range .Config.Env}}{{println .}}{{end}}'' | grep -E ''\\(TG_|TELEGRAM|ENABLED\\)''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -i ''logout\\\\|log.out\\\\|shutdown\\\\|stop'' | tail -20\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -E ''\\(getMe|getUpdates|sendMessage\\).*200'' | tail -5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -i ''success\\\\|started\\\\|初始化'' | head -20\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -E ''2026-03-\\(19|20|21\\)'' | grep -i ''error\\\\|fail\\\\|logout\\\\|400\\\\|401'' | head -20\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker stop clawbot && docker rm clawbot && echo ''✅ OpenClaw 已永久停用''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && docker-compose ps 2>/dev/null || ls -la docker-compose.yml 2>/dev/null || find /home/ollama -name ''docker-compose*.yml'' -type f 2>/dev/null | head -5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && docker-compose up -d && sleep 3 && docker-compose ps\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && docker compose up -d 2>&1 || docker run -d --name clawbot --restart unless-stopped -p 8088:8088 -v /var/run/docker.sock:/var/run/docker.sock 192.168.0.110:5000/library/clawbot:stable-v6 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot --tail=15 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps --format ''table {{.Names}}\\\\t{{.Status}}'' | grep -E ''clawbot|litellm''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && sed -i ''s|TELEGRAM_BOT_TOKEN=.*|TELEGRAM_BOT_TOKEN=8569720657:AAHrJ5CMOb4rP0IYJrCUiDViLsnpK69uEUI|'' .env && grep TELEGRAM_BOT_TOKEN .env\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && docker compose down && docker compose up -d && sleep 5 && docker logs clawbot --tail=10\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps --format ''{{.Names}}'' | grep -i alert\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker stop alertmanager && docker rm alertmanager && echo ''✅ 舊 AIOPS Alertmanager 已停用''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps --format ''table {{.Names}}\\\\t{{.Image}}\\\\t{{.Status}}''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cat /home/ollama/momo-pro/monitoring/prometheus/alert_rules.yml 2>/dev/null | grep -A5 ''ClawbotDown\\\\|telegram\\\\|AIOPS'' | head -30\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"find /home/ollama -name ''*.yml'' -type f 2>/dev/null | xargs grep -l ''ClawbotDown\\\\|telegram'' 2>/dev/null | head -5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec clawbot grep -r ''協同警報\\\\|ClawbotDown'' /app 2>/dev/null | head -5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec prometheus cat /etc/prometheus/prometheus.yml 2>/dev/null | grep -A10 ''alerting\\\\|alertmanager''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps | grep -i alert || echo ''✅ 沒有 alertmanager 在運行''\")",
-      "Bash(jq -r '.status, .components | to_entries[] | \"\"\"\"\\\\\\(.key\\): \\\\\\(.value.status\\)\"\"\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps --format ''table {{.Names}}\\\\t{{.Status}}'' | grep clawbot && docker logs clawbot --tail=15\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker inspect clawbot --format=''{{range .Config.Env}}{{println .}}{{end}}'' | grep TELEGRAM\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && sed -i ''s|TELEGRAM_BOT_TOKEN=.*|TELEGRAM_BOT_TOKEN=8569720657:AAFjDyjAN94QQrjn1gBnFXAyS20EUyozH8c|'' .env && docker compose down && docker compose up -d && sleep 5 && docker logs clawbot --tail=10\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec clawbot grep -r ''ClawBotDown\\\\|ClawbotDown'' /app 2>/dev/null | head -5 || echo ''在程式碼中找不到''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec prometheus cat /etc/prometheus/alerts.yml 2>/dev/null | grep -A10 ''ClawBot\\\\|clawbot'' | head -30\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec prometheus cat /etc/prometheus/alerts.yml 2>/dev/null | grep -i ''clawbot\\\\|claw'' -A5 -B5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot --since=5m 2>&1 | grep -i ''clawbot\\\\|incident\\\\|alert'' | tail -20\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot --tail 50 2>&1\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -i ''telegram\\\\|polling\\\\|bot'' | tail -20\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps --format ''table {{.Names}}\\\\t{{.Status}}\\\\t{{.Ports}}'' | grep -E ''claw|NAME''\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -E ''telegram|Telegram|error|Error'' | tail -20\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps | grep ollama\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps -a --format ''table {{.Names}}\\\\t{{.Status}}'' | head -20\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"sed -i ''s|host.docker.internal|172.17.0.1|g'' /home/ollama/clawbot-v5/.env && cat /home/ollama/clawbot-v5/.env | grep OLLAMA\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && docker-compose restart clawbot && sleep 3 && docker logs clawbot --tail 30 2>&1\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && docker compose restart clawbot && sleep 5 && docker logs clawbot --tail 30 2>&1\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec clawbot curl -s http://172.17.0.1:11434/api/tags | head -c 200\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | tail -10\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -iE ''error|telegram|polling|alert|send'' | tail -30\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cat /home/ollama/clawbot-v5/.env | grep OLLAMA\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd /home/ollama/clawbot-v5 && docker compose up -d --force-recreate clawbot && sleep 5 && docker logs clawbot 2>&1 | tail -20\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec clawbot curl -s http://172.17.0.1:11434/api/tags | head -c 100\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot --since 5m 2>&1 | tail -30\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec momo-db psql -U postgres -d clawbot -c \"\"SELECT enum_range\\(NULL::approvalstatus\\);\"\"\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec -e PGPASSWORD=clawbot123 momo-db psql -U clawbot -d clawbot -c \"\"SELECT enum_range\\(NULL::approvalstatus\\);\"\"\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps | grep -E ''postgres|db''\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker exec momo-db env | grep -i postgres\")",
-      "Bash(sshpass -p \"0936223270\" ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"PGPASSWORD=AwoooiProd2026 psql -h localhost -U awoooi -d awoooi_prod -c \"\"SELECT enum_range\\(NULL::approvalstatus\\);\"\"\")",
-      "Bash(KUBECONFIG=~/.kube/config kubectl config get-contexts)",
-      "Bash(docker tag:*)",
-      "Bash(docker push:*)",
-      "Bash(ssh ollama@192.168.0.188 \"cd ~/awoooi-build && find apps/web/src -name ''''*.ts'''' -o -name ''''*.tsx'''' | head -30 | xargs md5sum\")",
-      "Bash(rsync -avz --exclude 'node_modules' --exclude '.next' --exclude '.turbo' --exclude '*.log' /Users/ogt/awoooi/ ollama@192.168.0.188:~/awoooi-build/)",
-      "Bash(gh run:*)",
-      "Bash(APPROVAL_ID=\"ea43578e-17cd-40b9-b4c3-8fe8e92f225c\" __NEW_LINE_76dc92b2699cd7d5__ echo \"=== 檢查 Approval Metadata ===\" curl -s \"https://awoooi.wooo.work/api/v1/approvals/pending\")",
-      "Bash(APPROVAL_ID=\"865ab726-c3b9-447e-86a9-65a6227516e6\" __NEW_LINE_db14ef76ca26af32__ echo \"=== 簽核 ===\" curl -s -X POST \"https://awoooi.wooo.work/api/v1/approvals/$APPROVAL_ID/sign\" -H \"Content-Type: application/json\" -d '{\"\"\"\"signer_id\"\"\"\":\"\"\"\"commander\"\"\"\",\"\"\"\"signer_name\"\"\"\":\"\"\"\"Commander\"\"\"\",\"\"\"\"comment\"\"\"\":\"\"\"\"Test resolution\"\"\"\"}')",
-      "Read(//Users/ogt/awoooi/**)",
-      "Bash(APPROVAL_ID=\"e9445e68-6c3e-4899-b507-3b9b7bcaf0a7\" __NEW_LINE_680ad94d4896e58a__ echo \"=== 簽核 ===\" curl -s -X POST \"https://awoooi.wooo.work/api/v1/approvals/$APPROVAL_ID/sign\" -H \"Content-Type: application/json\" -d '{\"\"\"\"signer_id\"\"\"\":\"\"\"\"commander\"\"\"\",\"\"\"\"signer_name\"\"\"\":\"\"\"\"Commander\"\"\"\",\"\"\"\"comment\"\"\"\":\"\"\"\"Final test\"\"\"\"}')",
-      "Bash(APPROVAL_ID=\"eb0afb4e-834b-4af7-9ae0-3c58232fdd99\" INCIDENT=\"INC-20260323-F05CD6\" __NEW_LINE_47f1c3803a64b43c__ echo \"=== 簽核前 Incident 狀態 ===\" curl -s \"https://awoooi.wooo.work/api/v1/incidents/$INCIDENT\")",
-      "Bash(mkdir -p /Users/ogt/awoooi/.claude/hooks)",
-      "Bash(/Users/ogt/awoooi/.claude/hooks/pre-commit-check.sh:*)",
-      "Bash(git -C /Users/ogt/awoooi status packages/lewooogo-core/)",
-      "Bash(git -C /Users/ogt/awoooi ls-files packages/lewooogo-core/src/)",
-      "Bash(git -C /Users/ogt/awoooi status --short)",
-      "Bash(git -C /Users/ogt/awoooi add apps/api/pyproject.toml apps/api/scripts/ apps/api/src/ apps/web/.eslintrc.js apps/web/src/ packages/lewooogo-core/.eslintrc.js)",
-      "Bash(git -C /Users/ogt/awoooi diff --cached --stat)",
-      "Bash(git -C:*)",
-      "Bash(for wf:*)",
-      "Bash(do)",
-      "Bash(done)",
-      "Bash(jq 'if type == \"\"\"\"array\"\"\"\" then .[0] | {incident_id, status, decision} else . end')",
-      "Bash(PYTHONPATH=. python -c \"from src.api.v1.stats import router; print\\(''✅ stats.py 載入成功，路由數:'', len\\(router.routes\\)\\)\")",
-      "Bash(PYTHONPATH=. pytest tests/ -v --tb=short)",
-      "Bash(PYTHONPATH=. pytest tests/test_stats_api.py -v --tb=short)",
-      "Bash(PYTHONPATH=. pytest tests/test_webhook_telegram_integration.py::TestNewAlertTelegramPush -v --tb=long)",
-      "Bash(PYTHONPATH=. pytest tests/test_webhook_telegram_integration.py::TestNewAlertTelegramPush -v --tb=short)",
-      "Bash(PYTHONPATH=. pytest tests/test_webhook_telegram_integration.py -v --tb=short)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get pods -n awoooi')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get ns awoooi && kubectl get all -n awoooi')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get ns | head -20')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get pods -n awoooi-prod')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl logs awoooi-worker-bb89b5ffc-bpf45 -n awoooi-prod --tail=50')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl logs awoooi-worker-bb89b5ffc-bpf45 -n awoooi-prod --tail=100 | grep -i telegram')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl logs awoooi-api-8c9489b6c-cm8g5 -n awoooi-prod --tail=50 | grep -i webhook')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl logs awoooi-api-8c9489b6c-cm8g5 -n awoooi-prod --tail=30')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get pods -n monitoring | grep alertmanager')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"kubectl get configmap alertmanager-config -n monitoring -o jsonpath=''{.data.alertmanager\\\\.yml}'' | head -50\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get svc -n awoooi-prod')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"kubectl patch configmap alertmanager-config -n monitoring --type merge -p ''{\"\"data\"\":{\"\"alertmanager.yml\"\":\"\"global:\\\\n  resolve_timeout: 5m\\\\n\\\\nroute:\\\\n  group_by: [\\\\\"\"alertname\\\\\"\", \\\\\"\"severity\\\\\"\"]\\\\n  group_wait: 30s\\\\n  group_interval: 5m\\\\n  repeat_interval: 4h\\\\n  receiver: \\\\\"\"awoooi-webhook\\\\\"\"\\\\n  routes:\\\\n    - match:\\\\n        severity: critical\\\\n      receiver: \\\\\"\"awoooi-webhook\\\\\"\"\\\\n      group_wait: 10s\\\\n      repeat_interval: 1h\\\\n    - match:\\\\n        severity: warning\\\\n      receiver: \\\\\"\"awoooi-webhook\\\\\"\"\\\\n      group_wait: 1m\\\\n      repeat_interval: 4h\\\\n\\\\nreceivers:\\\\n  - name: \\\\\"\"awoooi-webhook\\\\\"\"\\\\n    webhook_configs:\\\\n      - url: \\\\\"\"http://192.168.0.120:32334/api/v1/webhook/alertmanager\\\\\"\"\\\\n        send_resolved: true\\\\n\\\\ninhibit_rules:\\\\n  - source_match:\\\\n      severity: \\\\\"\"critical\\\\\"\"\\\\n    target_match:\\\\n      severity: \\\\\"\"warning\\\\\"\"\\\\n    equal: [\\\\\"\"alertname\\\\\"\", \\\\\"\"instance\\\\\"\"]\\\\n\"\"}}''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl rollout restart deployment/alertmanager -n monitoring && kubectl rollout status deployment/alertmanager -n monitoring')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"kubectl get configmap alertmanager-config -n monitoring -o jsonpath=''{.data.alertmanager\\\\.yml}'' | grep -A 3 ''url:''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get pods -n awoooi-prod -o jsonpath=\"\"{range .items[*]}{.metadata.name}{\\\\\"\" \\\\\"\"}{.spec.containers[*].image}{\\\\\"\"\\\\\\\\n\\\\\"\"}{end}\"\"')",
-      "Bash(git mv:*)",
-      "Bash(for file:*)",
-      "Bash(do echo:*)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 wooo@192.168.0.120 \"echo ''Connected''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"kubectl get deployment -n awoooi-prod -o jsonpath=''{range .items[*]}{.metadata.name}{\"\" selector: \"\"}{.spec.selector.matchLabels}{\"\"\\\\n\"\"}{end}''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"kubectl delete deployment awoooi-api awoooi-web awoooi-worker -n awoooi-prod\")",
-      "WebFetch(domain:awoooi.wooo.work)",
-      "WebFetch(domain:api.awoooi.wooo.work)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get pods -n awoooi-prod -o wide')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get svc,ingress -n awoooi-prod')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl exec -n awoooi-prod deploy/awoooi-api -- curl -sf http://localhost:8000/api/v1/health 2>&1')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'curl -sf http://10.43.125.201:8000/api/v1/health 2>&1 || echo \"\"FAILED\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'sudo nginx -t 2>&1 && sudo cat /etc/nginx/sites-enabled/awoooi* 2>/dev/null || sudo cat /etc/nginx/conf.d/awoooi* 2>/dev/null || echo \"\"No awoooi nginx config found\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'cat /etc/nginx/sites-enabled/* 2>/dev/null | grep -A5 awoooi || cat /etc/nginx/conf.d/* 2>/dev/null | grep -A5 awoooi || ls -la /etc/nginx/ 2>/dev/null || echo \"\"No nginx on this host\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'ls /etc/nginx/sites-enabled/ 2>/dev/null && cat /etc/nginx/sites-enabled/*awoooi* 2>/dev/null || echo \"\"Checking conf.d...\"\" && ls /etc/nginx/conf.d/ 2>/dev/null')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'grep -l awoooi /etc/nginx/sites-enabled/* 2>/dev/null || grep -r \"\"awoooi\"\" /etc/nginx/sites-enabled/ 2>/dev/null | head -20')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'grep -r \"\"awoooi\\\\|32334\\\\|32335\"\" /etc/nginx/ 2>/dev/null | head -20')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S cp /tmp/awoooi-prod.conf /etc/nginx/conf.d/ && echo \"\"Config copied\"\" && sudo nginx -t 2>&1')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S ls -la /etc/nginx/ssl/ 2>/dev/null || echo \"\"No ssl dir\"\" && sudo ls -la /etc/letsencrypt/live/ 2>/dev/null | head -10')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S sed -i \"\"s|/etc/nginx/ssl/awoooi.crt|/etc/letsencrypt/live/awoooi.wooo.work/fullchain.pem|g\"\" /etc/nginx/conf.d/awoooi-prod.conf && sudo sed -i \"\"s|/etc/nginx/ssl/awoooi.key|/etc/letsencrypt/live/awoooi.wooo.work/privkey.pem|g\"\" /etc/nginx/conf.d/awoooi-prod.conf && echo \"\"Paths fixed\"\" && sudo nginx -t 2>&1')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S nginx -s reload && echo \"\"Nginx reloaded!\"\" && sleep 2')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'grep -r \"\"awoooi\"\" /etc/nginx/sites-enabled/ 2>/dev/null | head -5')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S grep -rl \"\"awoooi.wooo.work\"\" /etc/nginx/ 2>/dev/null')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'curl -sf http://192.168.0.121:32334/api/v1/health 2>&1 || echo \"\"FAILED to reach 121\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S rm /etc/nginx/conf.d/awoooi-prod.conf && sudo nginx -t && sudo nginx -s reload && echo \"\"Cleaned up duplicate config\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S tail -30 /var/log/nginx/error.log 2>/dev/null')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'grep -r \"\"api.awoooi\"\" /etc/nginx/ 2>/dev/null || echo \"\"No api.awoooi config found\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get configmap awoooi-config -n awoooi-prod -o yaml | grep -A5 NEXT_PUBLIC')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl get deployment awoooi-web -n awoooi-prod -o yaml | grep -A20 \"\"env:\"\" | head -25')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S tail -10 /var/log/nginx/access.log 2>/dev/null | grep awoooi')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S tail -5 /var/log/nginx/error.log 2>/dev/null')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S stat /etc/nginx/sites-available/awoooi.wooo.work.conf 2>/dev/null | grep -E \"\"Modify|Change|Birth\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl logs -n awoooi-prod -l app=awoooi-web --tail=30 2>/dev/null | grep -i \"\"api\\\\|error\\\\|fetch\"\" | head -20')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S tail -20 /var/log/nginx/access.log 2>/dev/null | grep -E \"\"awoooi.*api\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S tail -20 /var/log/nginx/awoooi-prod-access.log 2>/dev/null')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl exec -n awoooi-prod deploy/awoooi-web -- env | grep -i api')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl exec -n awoooi-prod deploy/awoooi-web -- sh -c \"\"grep -r \\\\\"\"NEXT_PUBLIC_API_URL\\\\|api.awoooi\\\\\"\" /app/.next/static/chunks/*.js 2>/dev/null | head -5 || grep -r \\\\\"\"awoooi.wooo.work\\\\\"\" /app/.next/static/chunks/*.js 2>/dev/null | head -3\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'kubectl exec -n awoooi-prod deploy/awoooi-web -- sh -c \"\"find /app/.next -name \\\\\"\"*.js\\\\\"\" -exec grep -l \\\\\"\"awoooi\\\\\"\" {} \\\\; 2>/dev/null | head -3\"\"')",
-      "Bash(./scripts/qa-zero-touch.sh)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S cat /etc/nginx/sites-available/awoooi.wooo.work.conf')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S cp /tmp/awoooi.wooo.work.conf /etc/nginx/sites-available/awoooi.wooo.work.conf && sudo nginx -t 2>&1')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'echo \"\"0936223270\"\" | sudo -S nginx -s reload && echo \"\"✅ Nginx reloaded with load balancing!\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'cd /opt && sudo ls -la sentry 2>/dev/null || echo \"\"Sentry 目錄不存在，需要建立\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'sudo mkdir -p /opt/sentry && sudo chown wooo:wooo /opt/sentry && cd /opt/sentry && git clone https://github.com/getsentry/self-hosted.git . 2>&1 | tail -5')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'echo \"\"0936223270\"\" | sudo -S mkdir -p /opt/sentry && echo \"\"0936223270\"\" | sudo -S chown wooo:wooo /opt/sentry && cd /opt/sentry && git clone https://github.com/getsentry/self-hosted.git . 2>&1 | tail -10')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'cd /opt/sentry && ls -la 2>&1 | head -20')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'cd /opt/sentry && git describe --tags 2>/dev/null || git rev-parse --short HEAD')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'cd /opt/sentry && ./install.sh --help 2>&1 | head -30 || echo \"\"No help available, checking script...\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'cd /opt/sentry && nohup ./install.sh --skip-user-creation --no-report-self-hosted-issues > /tmp/sentry-install.log 2>&1 &')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'tail -30 /tmp/sentry-install.log 2>/dev/null || echo \"\"日誌檔案尚未建立，等待中...\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'grep -E \"\"^\\\\▶|^Creating|^Starting|^Error|^✓|Pulling\"\" /tmp/sentry-install.log 2>/dev/null | tail -40')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'echo \"\"=== 日誌行數 ===\"\" && wc -l /tmp/sentry-install.log && echo \"\"\"\" && echo \"\"=== 最近進度 ===\"\" && tail -10 /tmp/sentry-install.log')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'echo \"\"=== 日誌行數 ===\"\" && wc -l /tmp/sentry-install.log && echo \"\"\"\" && echo \"\"=== 關鍵階段 ===\"\" && grep -E \"\"^▶|✓|Error|Creating|Starting\"\" /tmp/sentry-install.log | tail -20')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'echo \"\"=== 日誌行數 ===\"\" && wc -l /tmp/sentry-install.log && echo \"\"\"\" && echo \"\"=== 最近 20 行 ===\"\" && tail -20 /tmp/sentry-install.log')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'echo \"\"=== 日誌行數 ===\"\" && wc -l /tmp/sentry-install.log && echo \"\"\"\" && echo \"\"=== 關鍵階段 ===\"\" && grep -E \"\"^▶|✓|Error|Creating|Starting|Building|DONE\"\" /tmp/sentry-install.log | tail -30')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'echo \"\"=== 日誌行數 ===\"\" && wc -l /tmp/sentry-install.log && echo \"\"\"\" && echo \"\"=== 最近關鍵階段 ===\"\" && grep -E \"\"^▶|✓|Error|Creating|Starting|DONE|Completed|success\"\" /tmp/sentry-install.log | tail -25')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.110 'grep -E \"\"^▶|✓|Error|Completed|success|fail\"\" /tmp/sentry-install.log | tail -15')",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 KEYS incident:*)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cat /home/ollama/momo-pro/monitoring/alertmanager.yml 2>/dev/null || cat /etc/alertmanager/alertmanager.yml 2>/dev/null || echo ''Config not found''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot --tail 30 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker logs clawbot --tail 20 2>&1 | grep -iE ''telegram|send|alert|incident|error''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cat /home/ollama/clawbot-v5/.env | grep -E ''TELEGRAM|TG_'' | head -5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cat /home/ollama/clawbot-v5/.env | grep -E ''REDIS|POSTGRES|DATABASE'' | head -5\")",
-      "Bash(ssh ollama@192.168.0.188 'curl -s \"\"http://localhost:9093/api/v2/alerts?active=true\"\" | python3 -c \"\"import sys,json; alerts=json.load\\(sys.stdin\\); print\\(f\\\\\"\"Active alerts: {len\\(alerts\\)}\\\\\"\"\\)\"\"')",
-      "Bash(ssh ollama@192.168.0.188 'curl -s \"\"http://localhost:9093/api/v2/alerts\"\" | python3 -c \"\"import sys,json; alerts=json.load\\(sys.stdin\\); print\\(f\\\\\"\"Total alerts: {len\\(alerts\\)}\\\\\"\"\\); [print\\(a[\\\\\"\"labels\\\\\"\"][\\\\\"\"alertname\\\\\"\"]\\) for a in alerts[:5]]\"\"')",
-      "Bash(ssh ollama@192.168.0.188 'redis-cli -p 6380 -n 0 GET incident:INC-20260324-36AF55 | python3 -c \"\"import sys,json; d=json.load\\(sys.stdin\\); print\\(f\\\\\"\"Status: {d.get\\(\\\\\"\"status\\\\\"\"\\)}\\\\\"\"\\); print\\(f\\\\\"\"message_id: {d.get\\(\\\\\"\"message_id\\\\\"\", \\\\\"\"NONE\\\\\"\"\\)}\\\\\"\"\\); print\\(f\\\\\"\"chat_id: {d.get\\(\\\\\"\"chat_id\\\\\"\", \\\\\"\"NONE\\\\\"\"\\)}\\\\\"\"\\)\"\"')",
-      "Bash(ssh ollama@192.168.0.188 'redis-cli -p 6380 -n 0 GET incident:INC-20260324-36AF55 | python3 -c \"\"import sys,json; d=json.load\\(sys.stdin\\); print\\(f\\\\\"\"status: {d.get\\('status'\\)}\\\\\"\"\\); print\\(f\\\\\"\"message_id: {d.get\\('message_id'\\)}\\\\\"\"\\); print\\(f\\\\\"\"created_at: {d.get\\('created_at'\\)}\\\\\"\"\\)\"\"')",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 0 KEYS *approval*)",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 0 KEYS *incident*)",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 0 KEYS *pending*)",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 0 KEYS *)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/k3s-prod.yaml kubectl get pods -n awoooi-prod -o wide)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/k3s-prod.yaml kubectl get deployment awoooi-api -n awoooi-prod -o jsonpath='{.spec.template.spec.containers[0].image}')",
-      "Bash(kubectl --kubeconfig=/Users/ogt/awoooi/k3s-prod.yaml get deployment awoooi-api -n awoooi-prod -o jsonpath='{.spec.template.spec.containers[0].image}')",
-      "Bash(python3 -c \":*)",
-      "Bash(/tmp/awoooi-tg-secret.yaml:*)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/k3s-prod.yaml kubectl apply -f /tmp/awoooi-tg-secret.yaml)",
-      "Bash(for pod:*)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.188 \"curl -fsSL https://ollama.com/install.sh | sh\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no -o PreferredAuthentications=password wooo@192.168.0.188 \"echo connected && ollama --version\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no -o PreferredAuthentications=password ollama@192.168.0.188 \"curl -fsSL https://ollama.com/install.sh | sh\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S curl -fsSL https://ollama.com/install.sh | sudo -S sh\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"ollama --version\")",
-      "Bash(__NEW_LINE_95e9df111552805b__ echo:*)",
-      "Bash(sshpass -p '0936223270' scp /Users/ogt/awoooi/k8s/nginx/awoooi-prod.conf ollama@192.168.0.188:/tmp/awoooi-prod.conf)",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S cp /tmp/awoooi-prod.conf /etc/nginx/conf.d/awoooi-prod.conf && echo ''0936223270'' | sudo -S nginx -t 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S ls -la /etc/nginx/ssl/ 2>/dev/null || echo ''No ssl dir''; echo ''0936223270'' | sudo -S ls -la /etc/nginx/conf.d/ 2>/dev/null | head -10\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S grep -r ''ssl_certificate'' /etc/nginx/ 2>/dev/null | head -5\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S grep -A 20 ''server_name awoooi'' /etc/nginx/sites-enabled/all-sites.conf 2>/dev/null | head -30\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S ls -la /etc/nginx/sites-enabled/ 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S cat /etc/nginx/sites-available/awoooi.wooo.work.conf 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S rm /etc/nginx/conf.d/awoooi-prod.conf && echo ''0936223270'' | sudo -S nginx -t 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S nginx -s reload 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S systemctl reload nginx 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker logs openclaw 2>&1 | tail -30\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker ps -a --format ''table {{.Names}}\\\\t{{.Status}}\\\\t{{.Image}}'' 2>&1 | head -15\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -i telegram | tail -20\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker logs clawbot 2>&1 | tail -30\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker exec alertmanager cat /etc/alertmanager/alertmanager.yml 2>&1 | head -30\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"curl -sf ''http://localhost:9093/api/v1/alerts'' | jq ''.data | length'' 2>/dev/null || curl -sf ''http://localhost:9093/api/v2/alerts'' | jq ''length'' 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker exec alertmanager wget -qO- ''http://localhost:9093/api/v2/alerts'' 2>&1 | head -100\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"KUBECONFIG=/etc/rancher/k3s/k3s.yaml kubectl -n awoooi-prod logs -l app=awoooi-worker --tail=50 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"cat /home/ollama/alertmanager/alertmanager.yml 2>/dev/null || docker exec alertmanager cat /etc/alertmanager/alertmanager.yml\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker cp /tmp/alertmanager.yml alertmanager:/etc/alertmanager/alertmanager.yml && docker exec alertmanager amtool check-config /etc/alertmanager/alertmanager.yml && docker kill -s SIGHUP alertmanager\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker inspect alertmanager --format ''{{range .Mounts}}{{.Source}} -> {{.Destination}}{{println}}{{end}}''\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker exec alertmanager cat /etc/alertmanager/alertmanager.yml\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker restart alertmanager && sleep 3 && docker exec alertmanager cat /etc/alertmanager/alertmanager.yml\")",
-      "Bash(sshpass -p '0936223270' ssh ollama@192.168.0.188 \"docker logs clawbot 2>&1 | grep -i ''telegram\\\\|webhook\\\\|alert'' | tail -10\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=30 2>/dev/null | grep -E ''''POST|webhook|alertmanager|ManualTest''''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=30 2>/dev/null | grep -iE ''''POST|webhook''''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=50 2>/dev/null | grep -iE ''''POST.*webhook|alertmanager_webhook|NewFingerprint''''\")",
-      "Bash(kustomize build:*)",
-      "Bash(KUBECONFIG=~/.kube/config kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath='{.data}')",
-      "Bash(KUBECONFIG=/Users/ogt/.kube/config kubectl exec deploy/awoooi-api -n awoooi-prod -- env)",
-      "Bash(git checkout:*)",
-      "Bash(jq -r '.status // \"\"\"\"failed\"\"\"\"')",
-      "Bash(jq -r '.total // \"\"\"\"error\"\"\"\"')",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 10 XLEN awoooi:signals)",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 10 XRANGE awoooi:signals - + COUNT 5)",
-      "Bash(SENTRY_TOKEN=\"2b73050606d2b32f54095b4e177f4842f2bfe69d4b17da25f6daa4739148a972\" curl -s \"http://192.168.0.110:9000/api/0/organizations/\" -H \"Authorization: Bearer $SENTRY_TOKEN\")",
-      "Bash(SENTRY_TOKEN=\"2b73050606d2b32f54095b4e177f4842f2bfe69d4b17da25f6daa4739148a972\" curl -s \"http://192.168.0.110:9000/api/0/organizations/sentry/projects/\" -H \"Authorization: Bearer $SENTRY_TOKEN\")",
-      "Bash(SENTRY_TOKEN=\"2b73050606d2b32f54095b4e177f4842f2bfe69d4b17da25f6daa4739148a972\" curl -s \"http://192.168.0.110:9000/api/0/projects/sentry/awoooi-api/rules/\" -H \"Authorization: Bearer $SENTRY_TOKEN\")",
-      "Bash(SENTRY_TOKEN=\"2b73050606d2b32f54095b4e177f4842f2bfe69d4b17da25f6daa4739148a972\" __NEW_LINE_583db0bbb6875db0__ echo \"=== Alert Rules ===\" curl -s \"http://192.168.0.110:9000/api/0/projects/sentry/awoooi-api/rules/\" -H \"Authorization: Bearer $SENTRY_TOKEN\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get nodes -o wide && echo ''---'' && kubectl top nodes 2>/dev/null || echo ''metrics-server not installed''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod -o wide && echo ''---'' && kubectl get pvc -n awoooi-prod 2>/dev/null && echo ''---'' && kubectl get sc 2>/dev/null && echo ''---'' && kubectl get deploy -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get ns && echo ''---'' && kubectl get svc -A | grep -E ''prometheus|grafana|metrics|signoz|longhorn|argocd'' || echo ''No monitoring/gitops services found''\")",
-      "Bash(ssh wooo@192.168.0.120 \"cat /etc/rancher/k3s/config.yaml 2>/dev/null || echo ''--- K3s default config \\(no custom config.yaml\\) ---'' && echo ''---'' && sudo k3s check-config 2>/dev/null | head -30 || echo ''check-config not available''\")",
-      "Bash(ssh wooo@192.168.0.120 \"free -h && echo ''---'' && swapon --show && echo ''---'' && df -h /var/lib/rancher/k3s\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n cnpg-system && echo ''---'' && kubectl get svc -n monitoring\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get all -n awoooi-prod -o wide 2>/dev/null && echo ''---QUOTA---'' && kubectl describe quota -n awoooi-prod 2>/dev/null && echo ''---EVENTS---'' && kubectl get events -n awoooi-prod --sort-by=''.lastTimestamp'' 2>/dev/null | tail -20\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get helmcharts -A 2>/dev/null || echo ''No HelmCharts'' && echo ''---'' && kubectl get helmreleases -A 2>/dev/null || echo ''No HelmReleases'' && echo ''---'' && kubectl api-resources | grep -E ''argo|flux|velero|longhorn'' || echo ''No GitOps/Backup CRDs''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get ds -A && echo ''---'' && kubectl get cm -n kube-system | grep -E ''traefik|coredns'' && echo ''---REGISTRIES---'' && sudo cat /etc/rancher/k3s/registries.yaml 2>/dev/null || echo ''No registries.yaml''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get ingress -A 2>/dev/null || echo ''No Ingress'' && echo ''---HPA---'' && kubectl get hpa -A 2>/dev/null || echo ''No HPA'' && echo ''---PDB---'' && kubectl get pdb -A 2>/dev/null || echo ''No PDB'' && echo ''---SYSCTL---'' && cat /proc/sys/net/core/somaxconn && cat /proc/sys/fs/file-max\")",
-      "Bash(ssh wooo@192.168.0.120 \"systemctl status k3s | head -20 && echo ''---K3S-VERSION---'' && k3s --version && echo ''---ETCD-STATUS---'' && sudo k3s etcd-snapshot list 2>/dev/null | head -5 || echo ''No etcd snapshots''\")",
-      "Bash(ssh wooo@192.168.0.121 \"free -h && swapon --show && echo ''---DISK---'' && df -h /var/lib/rancher/k3s 2>/dev/null\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo ls -la /var/lib/rancher/k3s/server/db/ 2>/dev/null && echo ''---TOKEN---'' && sudo cat /var/lib/rancher/k3s/server/token 2>/dev/null | head -1 | cut -c1-20\")",
-      "Bash(ssh -o ConnectTimeout=10 wooo@192.168.0.120 \"ps aux | grep k3s | grep -v grep | head -3 && echo ''---'' && sudo cat /etc/systemd/system/k3s.service 2>/dev/null | grep -E ''ExecStart|datastore''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"echo ''0936223270'' | sudo -S mkdir -p /backup/k3s_etcd 2>/dev/null && echo ''0936223270'' | sudo -S chown ollama:ollama /backup/k3s_etcd 2>/dev/null && echo ''=== 188 備份目錄 ==='' && ls -la /backup/\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"mkdir -p ~/.ssh && chmod 700 ~/.ssh && echo ''ssh-rsa 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 wooo@mon'' >> ~/.ssh/authorized_keys && chmod 600 ~/.ssh/authorized_keys && echo ''SSH key 已加入''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"echo ''ssh-rsa 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 root@mon'' >> ~/.ssh/authorized_keys && echo ''Root SSH key added''\")",
-      "Bash(grep -r \"\"\"zod\"\"\" /Users/ogt/awoooi/package.json /Users/ogt/awoooi/apps/*/package.json /Users/ogt/awoooi/packages/*/package.json)",
-      "Bash(__NEW_LINE_144503b060dfd3dd__ echo:*)",
-      "Bash(__NEW_LINE_ae2a22b14586d7aa__ echo:*)",
-      "Bash(__NEW_LINE_e17561a4e55f74d4__ echo:*)",
-      "Bash(ssh wooo@192.168.0.120 \"echo ''''0936223270'''' | sudo -S cat /etc/rancher/k3s/k3s.yaml 2>/dev/null | sed ''''s|https://127.0.0.1:6443|https://192.168.0.125:6443|g''''\")",
-      "Bash(KUBECONFIG=/tmp/kubeconfig-vip.yaml kubectl get nodes)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml get rs -n awoooi-prod)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml get pods -A --no-headers)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml get jobs -A --no-headers)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml get rs -n awoooi-prod --no-headers)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml delete job api-watchdog-29556380 -n wooo-aiops-uat)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml get pods -n awoooi-prod)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml get pods -A)",
-      "Bash(kubectl --kubeconfig=/tmp/kubeconfig-vip.yaml get svc -A)",
-      "Bash(PGPASSWORD=changeme psql -h 192.168.0.188 -U awoooi -d awoooi_prod -f /Users/ogt/awoooi/apps/api/scripts/migrate_phase18_audit_logs.sql)",
-      "Bash(PLAYWRIGHT_BASE_URL=http://192.168.0.125:32335 npx playwright test phase11-conversational.spec.ts --reporter=list)",
-      "Bash(PLAYWRIGHT_BASE_URL=http://192.168.0.125:32335 npx playwright test phase11-conversational.spec.ts --reporter=list --workers=1)",
-      "Bash(KUBECONFIG=~/.kube/config kubectl get nodes --server=https://192.168.0.125:6443 --insecure-skip-tls-verify)",
-      "Bash(source .venv/bin/activate)",
-      "Read(//etc/postgresql/14/main/**)",
-      "Bash(for port:*)",
-      "Bash(kubectl top:*)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl top pods -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get pods -n awoooi-prod -o wide)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get svc -n awoooi-prod)",
-      "Bash(jq -r '.components | to_entries[] | \"\"\"\"\\\\\\(.key\\): \\\\\\(.value.status\\)\"\"\"\"')",
-      "Bash(tar -xzf velero-v1.13.0-darwin-arm64.tar.gz)",
-      "Bash(sudo mv:*)",
-      "Bash(velero version:*)",
-      "Bash(mkdir -p ~/bin)",
-      "Bash(mv velero-v1.13.0-darwin-arm64/velero ~/bin/)",
-      "Bash(~/bin/velero version:*)",
-      "Bash(k8s/velero/00-namespace.yaml:*)",
-      "Bash(k8s/velero/01-credentials.yaml:*)",
-      "Bash(k8s/velero/02-velero-install.yaml:*)",
-      "Bash(tar -xzf velero.tar.gz)",
-      "Bash(/tmp/velero-credentials:*)",
-      "Bash(__NEW_LINE_e85d95513fc16492__ ~/bin/velero install --provider aws --plugins velero/velero-plugin-for-aws:v1.9.0 --bucket velero-backups --secret-file /tmp/velero-credentials --backup-location-config region=minio,s3ForcePathStyle=true,s3Url=http://192.168.0.188:9000 --use-volume-snapshots=false --dry-run -o yaml)",
-      "Bash(__NEW_LINE_e85d95513fc16492__ head:*)",
-      "Bash(k8s/velero/README.md:*)",
-      "Bash(KUBECONFIG=/Users/ogt/.kube/config kubectl apply -f /Users/ogt/awoooi/k8s/velero/velero-install-full.yaml)",
-      "Bash(sshpass -p '09362233270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"whoami && hostname && cat /etc/sudoers.d/* 2>/dev/null | head -5 || echo ''no sudoers.d files''\")",
-      "Bash(sshpass -p '09362233270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"kubectl get nodes 2>&1 || echo ''kubectl failed, checking k3s kubeconfig...'' && ls -la /etc/rancher/k3s/k3s.yaml 2>&1\")",
-      "Bash(sshpass -p '09362233270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"sudo -l 2>&1 | head -20\")",
-      "Bash(sshpass -p '09362233270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''09362233270'' | sudo -S -l 2>&1\")",
-      "Bash(sshpass -p '09362233270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get nodes 2>&1\")",
-      "Bash(sshpass -p '0936223270' scp /Users/ogt/awoooi/k8s/velero/velero-install-full.yaml wooo@192.168.0.120:/tmp/velero-install-full.yaml)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''''0936223270'''' | sudo -S kubectl apply -f /tmp/velero-install-full.yaml 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get pods -n velero 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get backupstoragelocation -n velero 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl logs -n velero deploy/velero --tail=30 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl logs -n velero deploy/velero --tail=10 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get secret cloud-credentials -n velero -o jsonpath=''{.data.cloud}'' 2>&1 | base64 -d\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S curl -s http://192.168.0.188:9000/velero-backups/ 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl rollout restart deployment/velero -n velero 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get backups -n velero 2>&1\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl describe backup test-backup-20260328-2114 -n velero 2>&1 | tail -30\")",
-      "Bash(sshpass -p:*)",
-      "Read(//Users/ogt/awoooi/=== 測試 /approvals/**)",
-      "Bash(kubectl --kubeconfig=/Users/ogt/.kube/config get svc -n velero -o wide)",
-      "Bash(kubectl --kubeconfig=/Users/ogt/.kube/config get pods -n velero -o wide)",
-      "Bash(KUBECONFIG=/Users/ogt/.kube/config kubectl get svc -n velero)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'echo \"\"0936223270\"\" | sudo -S sh -c \"\"kubectl get pods -A | grep -E \\\\\"\"kube-state|state-metrics\\\\\"\"\"\"')",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'echo \"\"0936223270\"\" | sudo -S sh -c \"\"kubectl get ns | grep -E \\\\\"\"wooo|aiops|legacy|old\\\\\"\"\"\"')",
-      "Bash(KUBECONFIG=~/.kube/config kubectl get ns --no-headers)",
-      "WebFetch(domain:build.nvidia.com)",
-      "WebFetch(domain:ollama.com)",
-      "WebFetch(domain:docs.api.nvidia.com)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"curl -s ''http://admin:admin@localhost:3002/api/search?type=dash-db'' | python3 -c \"\"import sys,json; d=json.load\\(sys.stdin\\); print\\(f''Dashboard 數量: {len\\(d\\)}''\\); [print\\(f\\\\\"\"  - {i[''title'']}\\\\\"\"\\) for i in d[:10]]\"\"\")",
-      "Bash(jq '.ai_provider // .data.ai_provider // \"\"\"\"not found\"\"\"\"')",
-      "Bash(KUBECONFIG=~/.kube/config kubectl logs -n awoooi-prod deployment/awoooi-api --tail=50)",
-      "Bash(export NVIDIA_API_KEY=\"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\")",
-      "Bash(curl -s -X POST \"https://integrate.api.nvidia.com/v1/chat/completions\" -H \"Content-Type: application/json\" -H \"Authorization: Bearer $NVIDIA_API_KEY\" -d '{:*)",
-      "Bash(/tmp/fix-network-policy.yaml:*)",
-      "Bash(__NEW_LINE_acde7a92ceae01f6__ scp:*)",
-      "Bash(curl -s -X POST https://awoooi.wooo.work/api/v1/webhooks/alertmanager -H 'Content-Type: application/json' -d '{:*)",
-      "Bash(ssh ollama@192.168.0.188 'curl -s \"\"http://localhost:9090/api/v1/targets\"\" 2>/dev/null | grep -o \"\"\\\\\"\"health\\\\\"\":\\\\\"\"[^\\\\\"\"]*\\\\\"\"\"\" | sort | uniq -c')",
-      "Bash(ssh ollama@192.168.0.188 'curl -s \"\"http://localhost:9090/api/v1/rules\"\" 2>/dev/null | grep -o \"\"\\\\\"\"name\\\\\"\":\\\\\"\"[^\\\\\"\"]*\\\\\"\"\"\" | sort | uniq')",
-      "Bash(ssh ollama@192.168.0.188 'curl -s \"\"http://localhost:9090/api/v1/targets\"\" 2>/dev/null | grep -o \"\"\\\\\"\"job\\\\\"\":\\\\\"\"[^\\\\\"\"]*\\\\\"\"\"\" | sort | uniq -c | sort -rn')",
-      "Bash(ssh ollama@192.168.0.188 'curl -s \"\"http://localhost:9090/api/v1/query?query=up\"\" 2>/dev/null | grep -o \"\"\\\\\"\"instance\\\\\"\":\\\\\"\"[^\\\\\"\"]*\\\\\"\"\"\" | sort | uniq')",
-      "Bash(for i:*)",
-      "Bash(do sleep:*)",
-      "Bash(kubectl patch:*)",
-      "Bash(ssh wooo@192.168.0.110 \"cat /tmp/runner_clean.log 2>/dev/null; echo ''---''; ps aux | grep ''Runner.Listener'' | grep -v grep | wc -l\")",
-      "Bash(KUBECONFIG=~/.kube/config kubectl logs -n awoooi-prod -l app=awoooi-api --tail=200)",
-      "Bash(/Users/ogt/awoooi/ops/monitoring/deploy-exporters.sh:*)",
-      "WebFetch(domain:github.com)",
-      "WebFetch(domain:docs.ollama.com)",
-      "Skill(telegram:configure)",
-      "Skill(telegram:configure:*)",
-      "Bash(USE_NEW_ENGINE=true pytest tests/test_incident*.py -v --tb=short -x)",
-      "Bash(USE_NEW_ENGINE=true pytest tests/test_approval_field_alignment.py tests/test_learning_service.py -v --tb=short)",
-      "Bash(/tmp/debug_approval.py:*)",
-      "Bash(/tmp/debug_approval2.py:*)",
-      "Bash(/tmp/bulk_sign.sh:*)",
-      "Bash(bash /tmp/bulk_sign.sh)",
-      "Bash(/tmp/check_deploy.py:*)",
-      "Bash(/tmp/check_buttons.py:*)",
-      "Bash(ssh ollama@192.168.0.188 \"docker logs openclaw --since=10s 2>&1 | grep -Ev ''\\(GET|POST\\) /health'' | tail -10 && echo ''---'' && docker exec openclaw env | grep OPENAI_API_KEY | cut -c1-30\")",
-      "Read(//Users/ogt/awoooi/https:/awoooi.wooo.work/_next/static/chunks/app/%5Blocale%5D/**)",
-      "Bash(find /Users/ogt/awoooi/apps/web -type f \\\\\\(-name *.spec.ts -o -name *.spec.tsx \\\\\\))",
-      "Bash(kubectl -n awoooi-prod get pods)",
-      "Bash(kubectl -n production get pods)",
-      "Bash(ssh -o StrictHostKeyChecking=no wooo@192.168.0.121 \"export KUBECONFIG=/etc/rancher/k3s/k3s.yaml && sudo kubectl get deployment awoooi-web -n awoooi-prod -o jsonpath=''{.spec.template.spec.containers[0].image}'' && echo '''' && sudo kubectl get pods -n awoooi-prod -l app=awoooi-web --no-headers\")",
-      "Bash(KUBECONFIG=/Users/ogt/.kube/config kubectl get pods -n awoooi-prod)",
-      "Bash(for run_id in 166 165)",
-      "mcp__plugin_playwright_playwright__browser_navigate",
-      "mcp__plugin_playwright_playwright__browser_take_screenshot",
-      "Bash(open \"http://192.168.0.110:3001/wooo/awoooi/actions\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs?limit=5\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs/166/jobs\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs?limit=10\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runners\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/admin/runners\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs?limit=3\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs/169/jobs\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/jobs/179/logs\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" JOB_ID=180 curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/jobs/$JOB_ID/logs\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs?limit=2\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" JOB_ID=181 curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/jobs/$JOB_ID/logs\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs/172/jobs\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/jobs/182/logs\" -H \"Authorization: token $TOKEN\")",
-      "Bash(TOKEN=\"REDACTED_GITEA_TOKEN\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs/178\" -H \"Authorization: token $TOKEN\")",
-      "mcp__plugin_playwright_playwright__browser_snapshot",
-      "mcp__plugin_playwright_playwright__browser_fill_form",
-      "mcp__plugin_playwright_playwright__browser_click",
-      "Bash(GITEA_TOKEN=\"e6c9fecb1f0148939493ae0fa30407d28c91279d\" curl -s \"http://192.168.0.110:3001/api/v1/repos/wooo/awoooi/actions/runs?limit=5\" -H \"Authorization: token $GITEA_TOKEN\")",
-<<<<<<< Updated upstream
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 /tmp/a4_smoke.py)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from src.repositories.aider_event_repository import AiderEventRepository; print\\('import OK'\\)\")",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_service.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_service.py -v --tb=short)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from src.services.aider_event_service import classify_severity, should_create_incident, build_signal_data; print\\('✓ All imports successful'\\)\")",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_service.py::test_build_signal_data_redacts_secrets_in_annotations -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_events_api.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_service.py tests/test_aider_events_api.py tests/test_aider_event_models.py tests/test_secret_redactor.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_processor.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_processor.py tests/test_aider_event_service.py tests/test_aider_events_api.py tests/test_aider_event_models.py tests/test_secret_redactor.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from src.workers.aider_event_processor import AiderEventProcessor, get_aider_event_processor, run_aider_event_processor_loop; print\\('✓ All imports successful'\\)\")",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_processor.py -v --tb=short)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_processor.py tests/test_aider_event_service.py tests/test_aider_events_api.py tests/test_aider_event_models.py tests/test_secret_redactor.py --tb=short)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_ai_router_feedback.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_service.py tests/test_aider_events_api.py tests/test_aider_event_models.py tests/test_secret_redactor.py tests/test_aider_event_processor.py tests/test_ai_router_feedback.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from src.services.ai_router import AIRouter; from src.db.base import get_session_factory; print\\('✓ Imports successful, no circular imports'\\)\")",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_ai_router_feedback.py tests/test_aider_event_service.py -v --tb=short)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from src.api.v1 import aider_events; from src.workers.aider_event_processor import run_aider_event_processor_loop; from src.core.config import settings; print\\('AIDER_WEBHOOK_SECRET' in settings.__fields__, 'USE_AIDER_FEEDBACK' in settings.__fields__\\)\")",
-      "Bash(AIDER_WEBHOOK_SECRET=testsecret /Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from src.main import app; print\\('app OK; title:', app.title\\)\")",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_action_parsing.py tests/test_aider_event_service.py tests/test_aider_events_api.py tests/test_aider_event_models.py tests/test_secret_redactor.py tests/test_aider_event_processor.py tests/test_ai_router_feedback.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_action_parsing.py tests/test_aider_event_service.py tests/test_aider_events_api.py tests/test_aider_event_models.py tests/test_secret_redactor.py tests/test_aider_event_processor.py tests/test_ai_router_feedback.py -q)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pip install -e .[dev] --quiet)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pip install -e '.[dev]' --quiet)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/ -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from aider_watch_client.aiderw import main as awmain; from aider_watch_client.cli import main as climain; print\\('✓ imports ok'\\)\")",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pip show aider-watch-client)",
-      "Bash(tailscale status *)",
-      "Bash(kubectl rollout *)",
-      "Bash(bash /Users/ogt/awoooi/scripts/aider_watch_client/scripts/install.sh)",
-      "Bash(git rebase *)",
-      "Bash(/opt/homebrew/bin/aiderw --message \"add docstring to hello function\" --exit)",
-      "Bash(kubectl -n awoooi-prod get pod -l app=awoooi-api -o jsonpath='{.items[0].metadata.name}')",
-      "Bash(kubectl -n awoooi-prod exec awoooi-api-7b9464c969-8ml88 -- python -c ' *)",
-      "Bash(kubectl -n awoooi-prod rollout restart deployment/awoooi-api)",
-      "Bash(kubectl -n awoooi-prod get pod -l app=awoooi-api --no-headers)",
-      "Bash(kubectl -n awoooi-prod rollout status deployment/awoooi-api --timeout=120s)",
-      "Bash(/opt/homebrew/bin/aider-watch flush *)",
-      "Bash(kubectl -n awoooi-prod get pod -l app=awoooi-api -o wide)",
-      "Bash(kubectl -n awoooi-prod rollout status deployment/awoooi-api --timeout=30s)",
-      "Bash(kubectl -n awoooi-prod exec awoooi-api-6657fb9cf7-47lcg -- python -c \"import src.services.telegram_gateway as tg; import inspect; lines = inspect.getsource\\(tg\\); idx = lines.find\\('response_body=e.response.text'\\); print\\('FOUND' if idx >= 0 else 'NOT FOUND'\\)\")",
-      "Read(//opt/gitea/**)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/ -q)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/unit/test_aider_event_service.py tests/unit/test_aider_model.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_events_api.py tests/test_aider_event_models.py tests/test_aider_event_service.py tests/test_aider_event_processor.py -v)",
-      "Bash(kubectl -n awoooi-prod get svc)",
-      "Bash(kubectl -n openclaw get pod)",
-      "Bash(kubectl -n awoooi-prod exec awoooi-api-7cd784c875-r4qkz -- python -c ' *)",
-      "Bash(kubectl -n awoooi-prod logs awoooi-api-7cd784c875-qt6j2 --since=10m)",
-      "Bash(kubectl -n awoooi-prod logs awoooi-api-7cd784c875-qt6j2 --since=15m)",
-      "Bash(kubectl -n awoooi-prod logs awoooi-api-7cd784c875-qt6j2 --since=20m)",
-      "Bash(kubectl -n awoooi-prod get secret awoooi-secrets -o yaml)",
-      "Bash(kubectl -n awoooi-prod logs awoooi-api-7cd784c875-qt6j2 --since=30m)",
-      "Bash(kubectl -n awoooi-prod logs awoooi-api-7cd784c875-qt6j2 --since=2h)",
-      "Bash(kubectl -n awoooi-prod logs awoooi-api-7cd784c875-qt6j2)",
-      "Bash(kubectl -n awoooi-prod get pod -l app=awoooi-api -o jsonpath='{range .items[*]}{.metadata.name} {.status.containerStatuses[0].imageID}{\"\\\\n\"}{end}')",
-      "Bash(kubectl -n awoooi-prod get ingress)",
-      "Bash(kubectl -n awoooi-prod get svc awoooi-api-svc)",
-      "Bash(kubectl -n awoooi-prod logs -l app=awoooi-api --since=60s --prefix)",
-      "Bash(kubectl -n awoooi-prod logs -l app=awoooi-api --since=5m --prefix)",
-      "Bash(kubectl -n awoooi-prod logs pod/awoooi-api-86bc79766d-dn5ll --since=5m)",
-      "Bash(kubectl -n awoooi-prod logs pod/awoooi-api-86bc79766d-dn5ll --since=10m)",
-      "Bash(kubectl -n awoooi-prod logs pod/awoooi-api-86bc79766d-dn5ll)",
-      "Bash(kubectl -n awoooi-prod logs -l app=awoooi-api --since=90s --prefix)",
-      "Bash(kubectl -n awoooi-prod logs pod/awoooi-api-86bc79766d-4x69p --since=5m)",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 10 SCAN 0 MATCH \"playbook:PB-*\" COUNT 500)",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 -n 10 DBSIZE)",
-      "Bash(wait)",
-      "Read(//Users/**)",
-      "Read(//Users/ooo/.claude/**)",
-      "Bash(mkdir -p /Users/ogt/awoooi/.claude/agents)",
-      "Bash(cp /Users/ogt/.claude/agents/*.md /Users/ogt/awoooi/.claude/agents/)",
-      "Bash(kubectl -n awoooi-prod logs --tail=400 -l app=awoooi-api --prefix=true)",
-      "Bash(kubectl -n awoooi-prod logs --tail=300 awoooi-api-65c69fd649-bxbwp)",
-      "Bash(kubectl -n awoooi-prod logs --tail=20000 -l app=awoooi-api --prefix=false --since=24h)",
-      "Bash(kubectl -n awoooi-prod logs --since=24h awoooi-api-65c69fd649-bxbwp)",
-      "Bash(kubectl -n awoooi-prod logs --since=24h -l app=awoooi-api --prefix=false)",
-      "Bash(kubectl -n awoooi-prod logs --since=24h awoooi-api-65c69fd649-fmbxd)",
-      "Bash(kubectl -n awoooi-prod logs --since=3h awoooi-api-65c69fd649-fmbxd)",
-      "Bash(kubectl -n awoooi-prod logs --since=3h awoooi-api-65c69fd649-bxbwp)",
-      "Bash(kubectl -n awoooi-prod logs -l app=awoooi-api --tail=30 --since=30m)",
-      "Bash(kubectl -n awoooi-prod get pods -o wide)",
-      "Bash(kubectl -n awoooi-prod get pods -l app=awoooi-api -o jsonpath='{.items[0].metadata.creationTimestamp}')",
-      "Bash(kubectl -n awoooi-prod logs -l app=awoooi-api --tail=5 --since=5m)",
-      "Bash(kubectl -n awoooi-prod describe pod -l app=awoooi-api)",
-      "Bash(kubectl -n awoooi-prod logs -l app=awoooi-api --tail=20 --since=10m)",
-      "Bash(kubectl -n awoooi-prod exec deployment/awoooi-api -- python3 -c ' *)",
-      "Bash(PGPASSWORD=\"\" psql -h 188.188.188.188 -U aiops -d aiops -c \"\\\\d timeline_events\")",
-      "Bash(kubectl -n awoooi-prod get deploy awoooi-api -o yaml)",
-      "Bash(PGPASSWORD=\"\" psql --version)",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- env)",
-      "Bash(kubectl -n awoooi-prod logs --tail=500 deploy/awoooi-api)",
-      "Bash(kubectl cp *)",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'curl -sG \"$PROMETHEUS_URL/api/v1/query\" --data-urlencode \"query=up\" 2>&1 | head -c 400')",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'for q in \"sum\\(rate\\(http_requests_total{status=~\\\\\"5..\\\\\"}[5m]\\)\\) / sum\\(rate\\(http_requests_total[5m]\\)\\)\" \"avg\\(rate\\(container_cpu_usage_seconds_total{namespace=\\\\\"awoooi-prod\\\\\",container=\\\\\"awoooi-api\\\\\"}[5m]\\)\\)\" \"pg_stat_activity_count{datname=\\\\\"awoooi\\\\\"}\" \"increase\\(kube_pod_container_status_restarts_total{namespace=\\\\\"awoooi-prod\\\\\"}[15m]\\)\"; do echo \"---- $q\"; curl -sG \"$PROMETHEUS_URL/api/v1/query\" --data-urlencode \"query=$q\" 2>&1 | head -c 250; echo; done')",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'PGPASSWORD=as0V1mohktaFbGIx3R0iCatbMJ6XxFDL psql -h 192.168.0.188 -U awoooi -d awoooi_prod -c \"SELECT metric_name, count\\(*\\), max\\(trained_at\\) FROM dynamic_baseline_record GROUP BY metric_name;\" 2>&1 | head -20')",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'PGPASSWORD=as0V1mohktaFbGIx3R0iCatbMJ6XxFDL psql -h 192.168.0.188 -U awoooi -d awoooi_prod -c \"SELECT count\\(*\\) as asset_count FROM asset_inventory; SELECT count\\(*\\) as coverage_count FROM asset_coverage_snapshot; SELECT count\\(*\\) as host_cap_count FROM host_capacity_snapshot; SELECT count\\(*\\) as compl_count FROM asset_compliance_snapshot; SELECT count\\(*\\) as rule_cat FROM alert_rule_catalog; SELECT count\\(*\\) as log_cluster FROM log_cluster_record;\" 2>&1')",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'python3 -c \" *)",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- python3 -c ' *)",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'for q in \"http_requests_total\" \"container_cpu_usage_seconds_total\" \"container_memory_usage_bytes\" \"kube_pod_container_status_restarts_total\" \"pg_stat_activity_count\" \"node_cpu_seconds_total\" \"node_load1\"; do echo -n \"$q => \"; curl -sG \"$PROMETHEUS_URL/api/v1/query\" --data-urlencode \"query=count\\($q\\)\" 2>&1 | head -c 180; echo; done')",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'curl -sG \"$PROMETHEUS_URL/api/v1/query\" --data-urlencode \"query=container_cpu_usage_seconds_total\" 2>&1 | python3 -c \"import json,sys; d=json.load\\(sys.stdin\\); rs=d[\\\\\"data\\\\\"][\\\\\"result\\\\\"][:3]; [print\\(r[\\\\\"metric\\\\\"]\\) for r in rs]; print\\(\\\\\"total series:\\\\\", len\\(d[\\\\\"data\\\\\"][\\\\\"result\\\\\"]\\)\\)\"')",
-      "Bash(kubectl -n awoooi-prod exec deploy/awoooi-api -- sh -c 'which kubectl 2>&1; kubectl version --client 2>&1 | head -3; kubectl -n awoooi-prod get deploy awoooi-api 2>&1 | head -3')",
-      "Bash(kubectl -n awoooi-prod logs --tail=2000 deploy/awoooi-api)",
-      "Bash(psql --version)",
-      "WebFetch(domain:core.telegram.org)",
-      "mcp__plugin_context7_context7__resolve-library-id",
-      "mcp__plugin_context7_context7__query-docs",
-      "WebFetch(domain:docs.claude.com)",
-      "Bash(git tag *)",
-      "Read(//usr/**)",
-      "Bash(psql -h 192.168.0.110 -U awoooi_user -d awoooi -c \"SELECT id, alertname, status, confidence, description, created_at FROM approval_records WHERE status='PENDING' AND DATE\\(created_at AT TIME ZONE 'Asia/Taipei'\\) = CURRENT_DATE AT TIME ZONE 'Asia/Taipei' ORDER BY created_at DESC LIMIT 10;\")",
-      "Bash(kubectl -n awoooi-prod get deployment awoooi-api -o jsonpath='{.spec.template.spec.containers[0].image}')",
-      "Bash(kubectl -n awoooi-prod get deployment awoooi-api -o jsonpath='{.spec.template.spec.containers[0].imagePullPolicy}{\"\\\\n\"}{.spec.template.metadata.labels}{\"\\\\n\"}')",
-      "Bash(kubectl kustomize *)",
-      "Bash(kubectl -n awoooi-prod rollout status deployment/awoooi-api --timeout=60s)",
-      "Bash(kubectl -n awoooi-prod get pods -l app=awoooi-api --no-headers)",
-      "Bash(kubectl -n awoooi-prod patch deployment awoooi-api -p '{\"spec\":{\"template\":{\"spec\":{\"containers\":[{\"name\":\"api\",\"image\":\"192.168.0.110:5000/awoooi/api:cbd28e29a08435deb8c66af51654d8fa65120a14\"}]}}}}')",
-      "Bash(kubectl -n awoooi-prod get deployment awoooi-api -o jsonpath='{.spec.template.spec.containers[0].image}{\"\\\\n\"}')",
-      "Bash(kubectl -n awoooi-prod get pods -l app=awoooi-api -o jsonpath='{range .items[*]}{.metadata.name}{\"\\\\t\"}{.spec.containers[0].image}{\"\\\\n\"}{end}')",
-      "Bash(kubectl -n awoooi-prod get pdb awoooi-api-pdb -o jsonpath='{.spec.minAvailable}')",
-      "Bash(kubectl -n awoooi-prod get pods -l app=awoooi-api -o wide)",
-      "Bash(kubectl -n awoooi-prod describe rs -l app=awoooi-api)",
-      "Bash(kubectl -n awoooi-prod get events --sort-by='.lastTimestamp')",
-      "Bash(kubectl -n awoooi-prod get deployment awoooi-api -o jsonpath='{.spec.replicas}{\"\\\\n\"}{.status.replicas}{\"\\\\n\"}{.status.readyReplicas}{\"\\\\n\"}{.status.updatedReplicas}{\"\\\\n\"}')",
-      "Bash(kubectl -n awoooi-prod get pods -l app=awoooi-api --sort-by=.metadata.creationTimestamp -o jsonpath='{range .items[*]}{.metadata.name}{\":\"}{.metadata.creationTimestamp}{\"\\\\n\"}{end}')",
-      "Bash(kubectl -n awoooi-prod get deployment awoooi-api -o jsonpath='{.status.conditions[*]}')",
-      "Bash(kubectl -n awoooi-prod describe deployment awoooi-api)",
-      "Bash(kubectl -n awoooi-prod get rs -l app=awoooi-api -o jsonpath='{range .items[*]}{.metadata.name}{\":\"}{.spec.template.spec.containers[0].image}{\"\\\\n\"}{end}')",
-      "Bash(kubectl -n awoooi-prod get deployment awoooi-api -o yaml)",
-      "Bash(kubectl -n awoooi-prod rollout status deployment/awoooi-api --timeout=180s)",
-      "Bash(kubectl -n awoooi-prod set image deployment/awoooi-api api=192.168.0.110:5000/awoooi/api:cbd28e29a08435deb8c66af51654d8fa65120a14 --record=false)",
-      "Bash(kubectl -n awoooi-prod get pods -l app=awoooi-api -o jsonpath='{range .items[*]}{.metadata.name}{\"\\\\t\"}{.spec.containers[0].image}{\"\\\\t\"}{.status.phase}{\"\\\\n\"}{end}')",
-      "Bash(kubectl -n awoooi-prod get deployment awoooi-api -o jsonpath='{.status.replicas}{\"\\\\t\"}{.status.readyReplicas}{\"\\\\t\"}{.status.updatedReplicas}')",
-      "Bash(bash /tmp/diagnostic.sh)",
-      "WebFetch(domain:docs.github.com)",
-      "WebFetch(domain:docs.sonarsource.com)",
-      "WebFetch(domain:gitea.com)",
-      "WebFetch(domain:docs.gitea.com)",
-      "WebFetch(domain:www.sonarsource.com)",
-      "WebFetch(domain:golangci-lint.run)",
-      "WebFetch(domain:www.uber.com)",
-      "Bash(bash scripts/ops/deploy-alerts.sh --dry-run)",
-      "Bash(bash scripts/ops/deploy-alerts.sh)",
-      "Bash(promtool check *)",
-      "WebFetch(domain:openrouter.ai)",
-      "WebFetch(domain:qwenlm.github.io)",
-      "WebFetch(domain:aclanthology.org)",
-      "WebFetch(domain:datanorth.ai)",
-      "WebFetch(domain:www.infoq.com)",
-      "WebFetch(domain:aws.amazon.com)",
-      "WebFetch(domain:artificialanalysis.ai)",
-      "WebFetch(domain:www.alibabacloud.com)",
-      "WebFetch(domain:docs.langchain.com)",
-      "WebFetch(domain:arxiv.org)",
-      "WebFetch(domain:blog.kilo.ai)",
-      "WebFetch(domain:www.siliconflow.com)",
-      "WebFetch(domain:aicompetence.org)",
-      "Bash(redis-cli -h 192.168.0.188 -p 6380 ping)",
-      "Bash(redis-cli ping *)"
-=======
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest apps/api/tests/test_aider_event_models.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_action_parsing.py -v --collect-only)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_action_parsing.py --collect-only)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -m pytest tests/test_aider_event_models.py tests/test_secret_redactor.py -v)",
-      "Bash(/Users/ogt/.pyenv/versions/3.11.7/bin/python3 -c \"from src.repositories.aider_event_repository import AiderEventRepository; print\\('import OK'\\)\")"
->>>>>>> Stashed changes
-    ],
-    "deny": [
-      "Bash(rm -rf *)",
-      "Bash(git push --force *)",
-      "Bash(git reset --hard *)",
-      "Bash(kubectl delete *)",
-      "Bash(docker rm -f *)"
-    ],
-    "additionalDirectories": [
-      "/Users/ogt/.claude/projects/-Users-ogt-awoooi/memory",
-      "/Users/ogt/awoooi/.claude/hooks",
-      "/Users/ogt/.claude/channels/telegram",
-<<<<<<< Updated upstream
-      "/Users/ogt",
-      "/Users/ogt/.claude",
-      "/Users/ogt/awoooi/apps/web/src/app/[locale]/aiops"
-    ]
-  },
-  "hooks": {
-    "PreToolUse": [
-      {
-        "matcher": "",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "node $CLAUDE_PROJECT_DIR/.claude/hooks/awoooi-guard.js 2>/dev/null || true"
-          },
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/branch-protection.js"
-          },
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/commit-quality.js"
-          },
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/large-file-warner.js"
-          },
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/mcp-health.js"
-          }
-        ]
-      }
-    ],
-    "PostToolUse": [
-      {
-        "matcher": "",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/audit-log.js"
-          },
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/suggest-compact.js"
-          }
-        ]
-      }
-    ],
-    "Stop": [
-      {
-        "matcher": "",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/cost-tracker.js"
-          },
-          {
-            "type": "command",
-            "command": "node /Users/ogt/.claude/hooks/session-summary.js"
-          }
-        ]
-      }
-=======
-      "/Users/ogt/aider-watch"
->>>>>>> Stashed changes
-    ]
-  }
-}

.claude/settings.json.bak.20260323

@@ -1,827 +0,0 @@
-{
-  "permissions": {
-    "allow": [
-      "Bash(pnpm install:*)",
-      "Bash(npm --version)",
-      "Bash(npm install:*)",
-      "Bash(pnpm --version)",
-      "Bash(pnpm dev:*)",
-      "Bash(pnpm add:*)",
-      "Bash(ls -la /Users/ogt/awoooi/apps/web/next.config.*)",
-      "Bash(pkill -f \"next dev\")",
-      "Bash(curl -sL http://localhost:3000/zh-TW)",
-      "Bash(curl -s http://localhost:3000/zh-TW)",
-      "Bash(pnpm --filter web build)",
-      "Bash(curl -s http://localhost:3001/zh-TW)",
-      "Bash(curl -s -o /dev/null -w \"%{http_code}\" http://localhost:3000/zh-TW)",
-      "Bash(kubectl apply:*)",
-      "Bash(chmod +x /Users/ogt/awoooi/deploy-infra.sh)",
-      "Bash(./deploy-infra.sh)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"mkdir -p /tmp/awoooi-k8s\")",
-      "Bash(sshpass -p '0936223270' scp -o StrictHostKeyChecking=no /Users/ogt/awoooi/k8s/awoooi-prod/01-namespace-quota.yaml /Users/ogt/awoooi/k8s/awoooi-prod/02-network-policy.yaml /Users/ogt/awoooi/k8s/awoooi-prod/04-configmap.yaml wooo@192.168.0.120:/tmp/awoooi-k8s/)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"sudo kubectl apply -f /tmp/awoooi-k8s/01-namespace-quota.yaml\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl apply -f /tmp/awoooi-k8s/01-namespace-quota.yaml 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl apply -f /tmp/awoooi-k8s/02-network-policy.yaml 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl apply -f /tmp/awoooi-k8s/04-configmap.yaml 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get ns awoooi-prod -o wide 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get networkpolicy -n awoooi-prod 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get resourcequota,limitrange,configmap -n awoooi-prod 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"rm -rf /tmp/awoooi-k8s\")",
-      "Bash(PYTHONPATH=. python -c \"from src.main import app; print\\(''Import OK''\\)\")",
-      "Bash(curl -s http://localhost:8000/api/v1/health/ready)",
-      "Bash(curl -s http://localhost:8000/api/v1/health/live)",
-      "Bash(curl -s http://localhost:8000/)",
-      "Bash(pkill -f \"uvicorn src.main:app\")",
-      "Bash(pkill -f \"node.*next\")",
-      "Bash(curl -s http://localhost:8000/api/v1/health)",
-      "Read(//Users/ogt/awoooi/apps/api/**)",
-      "Bash(pnpm typecheck:*)",
-      "Read(//Users/ogt/awoooi/apps/web/**)",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/dashboard/demo/spike/clear)",
-      "Read(//Users/ogt/awoooi/=== 驗證英文頁面 \\(/en/**)",
-      "Bash(jq \".devDependencies | keys | map\\(select\\(startswith\\(\"\"@playwright\"\"\\) or startswith\\(\"\"playwright\"\"\\)\\)\\)\")",
-      "Bash(npx playwright:*)",
-      "Bash(curl -s http://localhost:3000/zh-TW/demo -o /dev/null -w \"Frontend: HTTP %{http_code}\\\\n\")",
-      "Bash(__NEW_LINE_ef548029029cdfac__ echo:*)",
-      "Bash(curl -s http://localhost:8000/api/v1/health -o /dev/null -w \"Backend: HTTP %{http_code}\\\\n\")",
-      "Bash(echo '=== 已產出的截圖 ===' find /Users/ogt/awoooi/apps/web/test-results -name *.png)",
-      "Bash(echo '=== Playwright E2E 測試結果 ===' echo echo '📸 截圖證據 \\(test-results/screenshots/\\):' ls -la /Users/ogt/awoooi/apps/web/test-results/screenshots/ __NEW_LINE_db74e5f56e34db17__ echo echo '🎬 錄影證據 \\(.webm\\):' find /Users/ogt/awoooi/apps/web/test-results -name *.webm -exec ls -la {})",
-      "Bash(__NEW_LINE_db74e5f56e34db17__ echo:*)",
-      "Bash(source .venv/bin/activate)",
-      "Bash(python scripts/demo_multisig.py)",
-      "Bash(python -c \"from src.api.v1.approvals import router; print\\(''✅ Approvals router loaded:'', len\\(router.routes\\), ''routes''\\)\")",
-      "Bash(npx tsc:*)",
-      "Bash(chmod +x /Users/ogt/awoooi/scripts/demo-multisig-flow.sh)",
-      "Bash(python -c \"from src.main import app; print\\(''✅ API loads successfully''\\)\")",
-      "Bash(jq)",
-      "Bash(/Users/ogt/awoooi/scripts/demo-multisig-flow.sh)",
-      "Bash(curl -s -X POST \"http://localhost:8000/api/v1/approvals\" -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(curl -s http://localhost:8000/api/v1/openapi.json)",
-      "Bash(python -c \":*)",
-      "Bash(curl -s http://localhost:3000 -o /dev/null -w \"%{http_code}\")",
-      "Bash(lsof -ti:3000,3001,8000)",
-      "Bash(curl -s http://localhost:8000/health)",
-      "Bash(curl -s http://localhost:8000/api/v1/approvals/pending)",
-      "Bash(curl -s -o /dev/null -w \"%{http_code}\" http://localhost:3001/zh-TW/demo)",
-      "Bash(ls -la test-results/*.png)",
-      "Bash(cp test-results/cpo102-*.png /Users/ogt/awoooi/docs/screenshots/)",
-      "Bash(ssh ogt@192.168.0.120 'cat /etc/rancher/k3s/k3s.yaml')",
-      "Bash(python -c \"from src.main import app; print\\(''✅ main.py imports OK''\\)\")",
-      "Bash(curl -s http://localhost:8000/api/v1/approvals/k8s-test)",
-      "Bash(sqlite3 awoooi.db \".tables\")",
-      "Bash(sshpass -p 0936223270 ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 'sudo cat /etc/rancher/k3s/k3s.yaml')",
-      "Bash(kubectl --kubeconfig=/Users/ogt/awoooi/apps/api/k3s-prod.yaml get deployments -n awoooi-prod)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get deployments -n awoooi-prod 2>/dev/null\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get deployments -A 2>/dev/null\")",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/approvals -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(APPROVAL_ID=\"b58a0d86-fa4e-43ca-881c-02e978cd7943\")",
-      "Bash(curl -s -X POST \"http://localhost:8000/api/v1/approvals/$APPROVAL_ID/sign\" -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(sqlite3 /Users/ogt/awoooi/apps/api/awoooi.db \"SELECT operation_type, target_resource, namespace, success, dry_run_passed, dry_run_message, error_message, execution_duration_ms, created_at FROM audit_logs ORDER BY created_at DESC LIMIT 1;\" -header -column)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S kubectl get pods -n monitoring -l app=grafana 2>/dev/null\")",
-      "Bash(curl -s http://192.168.0.188:11434/api/tags)",
-      "Bash(python -c \"from src.main import app; print\\(''✅ Compile OK''\\)\")",
-      "Bash(curl -s http://localhost:8000/api/v1/ai/status)",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/ai/analyze-and-propose -H \"Content-Type: application/json\" -d '{}')",
-      "Bash(curl -s -X POST http://192.168.0.188:11434/api/generate -H \"Content-Type: application/json\" -d '{\"\"\"\"model\"\"\"\":\"\"\"\"llama3.2:1b\"\"\"\",\"\"\"\"prompt\"\"\"\":\"\"\"\"Output only JSON: {\\\\\"\"\"\"action\\\\\"\"\"\":\\\\\"\"\"\"test\\\\\"\"\"\"}\"\"\"\",\"\"\"\"stream\"\"\"\":false,\"\"\"\"format\"\"\"\":\"\"\"\"json\"\"\"\"}' --max-time 30)",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/ai/analyze-and-propose -H \"Content-Type: application/json\" -d '{}' --max-time 60)",
-      "Bash(PROMPT='你是 ClawBot AI。分析以下監控數據，輸出純 JSON（無其他文字）。:*)",
-      "Bash(curl -s -X POST http://192.168.0.188:11434/api/generate -H \"Content-Type: application/json\" -d \"{\"\"model\"\":\"\"llama3.2:1b\"\",\"\"prompt\"\":\"\"$PROMPT\"\",\"\"stream\"\":false,\"\"format\"\":\"\"json\"\",\"\"options\"\":{\"\"num_predict\"\":256,\"\"temperature\"\":0.1}}\" --max-time 60)",
-      "Bash(curl -s -X POST http://192.168.0.188:11434/api/generate -H \"Content-Type: application/json\" -d '{\"\"\"\"model\"\"\"\":\"\"\"\"llama3.2:1b\"\"\"\",\"\"\"\"prompt\"\"\"\":\"\"\"\"Harbor service returning 404. Output JSON: {\\\\\"\"\"\"suggested_action\\\\\"\"\"\":\\\\\"\"\"\"RESTART_DEPLOYMENT\\\\\"\"\"\",\\\\\"\"\"\"target_resource\\\\\"\"\"\":\\\\\"\"\"\"harbor\\\\\"\"\"\",\\\\\"\"\"\"namespace\\\\\"\"\"\":\\\\\"\"\"\"default\\\\\"\"\"\",\\\\\"\"\"\"risk_level\\\\\"\"\"\":\\\\\"\"\"\"medium\\\\\"\"\"\",\\\\\"\"\"\"reasoning\\\\\"\"\"\":\\\\\"\"\"\"Service down\\\\\"\"\"\",\\\\\"\"\"\"confidence\\\\\"\"\"\":0.8,\\\\\"\"\"\"affected_services\\\\\"\"\"\":[]}\"\"\"\",\"\"\"\"stream\"\"\"\":false,\"\"\"\"format\"\"\"\":\"\"\"\"json\"\"\"\",\"\"\"\"options\"\"\"\":{\"\"\"\"num_predict\"\"\"\":128,\"\"\"\"temperature\"\"\"\":0.1}}' --max-time 30)",
-      "Bash(curl -v -X POST http://192.168.0.188:11434/api/generate -H \"Content-Type: application/json\" -d '{\"\"\"\"model\"\"\"\":\"\"\"\"llama3.2:1b\"\"\"\",\"\"\"\"prompt\"\"\"\":\"\"\"\"Say hello\"\"\"\",\"\"\"\"stream\"\"\"\":false}' --max-time 30)",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/ai/analyze-and-propose -H \"Content-Type: application/json\" -d '{}' --max-time 120)",
-      "Bash(curl -s http://localhost:8000/api/v1/ai/analyze-and-propose -X POST -H \"Content-Type: application/json\")",
-      "Bash(curl -s http://localhost:8000/api/v1/dashboard)",
-      "Bash(ls -la ~/Downloads/image*.png)",
-      "Bash(ls -la ~/Desktop/image*.png)",
-      "Bash(ls -la /Users/ogt/awoooi/apps/web/public/*.png)",
-      "WebFetch(domain:openclaw.ai)",
-      "Bash(ls -la /Users/ogt/Downloads/*.png)",
-      "Bash(ls -la /Users/ogt/.gemini/antigravity/brain/*/image*.png)",
-      "Bash(ls -lat /Users/ogt/Downloads/*.png)",
-      "Bash(curl -s http://localhost:8000/api/v1/approvals)",
-      "Bash(curl -s -X GET http://localhost:8000/api/v1/approvals/)",
-      "Bash(APPROVAL_ID=\"4989729e-e518-4e7e-8dff-5c3269e0c82b\")",
-      "Bash(curl -s -X POST \"http://localhost:8000/api/v1/approvals/$APPROVAL_ID/sign\" -H \"Content-Type: application/json\" -d '{\"\"\"\"signer_id\"\"\"\": \"\"\"\"ciso-001\"\"\"\", \"\"\"\"signer_name\"\"\"\": \"\"\"\"Demo CISO\"\"\"\", \"\"\"\"comment\"\"\"\": \"\"\"\"資安確認，核准執行\"\"\"\"}')",
-      "Bash(curl -s http://localhost:8000/api/v1/webhooks/health)",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/webhooks/alerts -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(curl -s http://localhost:3000)",
-      "Bash(ls -la apps/web/test-results/*.png)",
-      "Bash(curl -s http://localhost:3000/zh-TW/demo)",
-      "Bash(curl -s -o /dev/null -w \"%{http_code}\" http://localhost:3333/zh-TW/demo)",
-      "Bash(curl -s http://localhost:8001/api/v1/approvals/pending)",
-      "Bash(curl -s -X POST http://localhost:8001/api/v1/approvals -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(curl -s http://localhost:8001/openapi.json)",
-      "Bash(curl -s http://localhost:8001/docs)",
-      "Bash(curl -s http://localhost:8001/api/v1/webhooks/grafana -X OPTIONS)",
-      "Bash(pnpm run:*)",
-      "Bash(node scripts/screenshot-rbac.mjs)",
-      "Bash(pnpm exec:*)",
-      "Bash(curl -s http://localhost:3333 -o /dev/null -w \"%{http_code}\")",
-      "Bash(curl -s http://localhost:3333/zh-TW/demo -o /dev/null -w \"%{http_code}\")",
-      "Bash(python3 -c \"import sys,json; d=json.load\\(sys.stdin\\); print\\(f''''Count: {d[count]}''''\\); [print\\(f''''- {a[id][:8]}... risk={a[risk_level]}''''\\) for a in d[''''approvals''''][:3]]\")",
-      "Bash(curl -s http://localhost:3000/zh-TW/demo -o /dev/null -w \"%{http_code}\")",
-      "Bash(python -c \"import sys,json; d=json.load\\(sys.stdin\\); print\\(f''''   Connected: {d[\"\"success\"\"]}''''\\); print\\(f''''   Namespaces: {d[\"\"namespaces\"\"][:3]}...''''\\)\" __NEW_LINE_57ae1c1c812968e7__ echo \"\" echo \"3. 資料庫持久化:\" sqlite3 /Users/ogt/awoooi/apps/api/awoooi.db \"SELECT COUNT\\(*\\) as approvals FROM approval_records;\" sqlite3 /Users/ogt/awoooi/apps/api/awoooi.db \"SELECT COUNT\\(*\\) as timeline FROM timeline_events;\" sqlite3 /Users/ogt/awoooi/apps/api/awoooi.db \"SELECT COUNT\\(*\\) as audits FROM audit_logs;\")",
-      "Bash(head -2 __NEW_LINE_9bf9481fbdf30d4e__ echo \"\" echo \"2. 告警收斂跳過 LLM 日誌 \\(應該有 4 次\\):\" grep -c \"alert_converged_skip_llm\" /tmp/api-server.log)",
-      "Bash(python -m json.tool)",
-      "Bash(__NEW_LINE_7463bff94cecc20f__ echo:*)",
-      "Bash(__NEW_LINE_13846c8488c5fa9a__ echo:*)",
-      "Bash(__NEW_LINE_13846c8488c5fa9a__ ls:*)",
-      "Bash(python -c \"import sys,json; d=json.load\\(sys.stdin\\); print\\(f''''   Status: {d[\"\"status\"\"]}''''\\)\" __NEW_LINE_32366ca1bb050259__ echo \"\" echo \"2. 待簽核記錄 \\(含 hit_count\\):\" curl -s http://localhost:8000/api/v1/approvals/pending)",
-      "Read(//Users/ogt/awoooi/**)",
-      "Bash(curl -s http://localhost:8000/api/v1/timeline/events?limit=10)",
-      "Bash(curl -s http://localhost:8000/api/v1/timeline/events?limit=5)",
-      "Bash(ls -la /Users/ogt/awoooi/apps/api/*.txt /Users/ogt/awoooi/apps/api/*.toml)",
-      "Bash(ls -la /Users/ogt/awoooi/docker-compose*.yml)",
-      "Bash(ls /Users/ogt/awoooi/k8s/awoooi-prod/*rbac* /Users/ogt/awoooi/k8s/awoooi-prod/*service-account*)",
-      "Bash(kubectl kustomize:*)",
-      "Bash(docker compose:*)",
-      "Bash(docker info:*)",
-      "Bash(python3 -c \"import sys,json; d=json.load\\(sys.stdin\\); print\\(''''API Status:'''', d.get\\(''''status'''', ''''unknown''''\\)\\)\")",
-      "Bash(pkill -9 -f uvicorn)",
-      "Bash(lsof -ti:8000)",
-      "Bash(open -a Docker)",
-      "Bash(docker stop:*)",
-      "Bash(lsof -ti:3000)",
-      "Bash(docker start:*)",
-      "Bash(docker ps:*)",
-      "Bash(curl -s http://localhost:3000 -o /dev/null -w 'HTTP Status: %{http_code}\\\\n')",
-      "Bash(curl -I http://localhost:8000/api/v1/dashboard/stream)",
-      "Bash(curl -s http://localhost:8000/openapi.json)",
-      "Bash(curl -s http://localhost:8000/api/v1/dashboard/stream --max-time 3 -w \"\\\\n--- HTTP Status: %{http_code} ---\\\\n\")",
-      "Bash(curl -s http://localhost:8000/api/v1/dashboard/stream --max-time 3)",
-      "Bash(curl -s http://localhost:3000/zh-TW -o /dev/null -w \"HTTP Status: %{http_code}\\\\n\")",
-      "Bash(curl -s -D - http://localhost:8000/api/v1/dashboard/stream --max-time 2)",
-      "Bash(chmod +x /Users/ogt/awoooi/scripts/deploy-infra.sh)",
-      "Bash(./scripts/deploy-infra.sh)",
-      "Bash(pnpm --filter @awoooi/web build)",
-      "Bash(timeout 10 env MOCK_MODE=true OTEL_ENABLED=false uvicorn src.main:app --host 0.0.0.0 --port 8099)",
-      "Bash(timeout 8 pnpm --filter @awoooi/web dev)",
-      "Bash(git diff:*)",
-      "Bash(curl -s -I http://localhost:8000/api/v1/dashboard/stream)",
-      "Bash(timeout 3 curl -s -N http://localhost:8000/api/v1/dashboard/stream)",
-      "Bash(grep -n \"NEXT_PUBLIC\\\\|API_URL\\\\|localhost\" /Users/ogt/awoooi/apps/web/.env*)",
-      "Bash(timeout 2 curl -s -D - -N http://localhost:8000/api/v1/dashboard/stream)",
-      "Bash(curl -s http://localhost:3000/)",
-      "Bash(python -m py_compile scripts/fire_test_alert.py)",
-      "Bash(python -m scripts.fire_test_alert --help)",
-      "Bash(python -m scripts.fire_test_alert)",
-      "Bash(python -m scripts.fire_test_alert --type k8s_pod_crash)",
-      "Bash(timeout 3 curl -s -N -H \"Origin: http://localhost:3000\" http://localhost:8000/api/v1/dashboard/stream)",
-      "Bash(python -m scripts.fire_test_alert --type disk_full)",
-      "Bash(docker restart:*)",
-      "Bash(curl -s -w \"\\\\nHTTP_CODE: %{http_code}\\\\n\" http://localhost:3000)",
-      "Bash(docker exec:*)",
-      "Bash(docker rmi:*)",
-      "Bash(timeout 5 curl -s -N http://localhost:8000/api/v1/dashboard/stream)",
-      "Bash(curl -s http://localhost:3000 -w \"\\\\nHTTP: %{http_code}\\\\n\")",
-      "Bash(timeout 120 docker logs awoooi-api -f --since 1s)",
-      "Bash(curl -s -I -H \"Origin: http://localhost:3000\" http://localhost:8000/api/v1/dashboard/stream)",
-      "Bash(curl -s -X OPTIONS -H \"Origin: http://localhost:3000\" -H \"Access-Control-Request-Method: GET\" http://localhost:8000/api/v1/dashboard/stream -I)",
-      "Bash(node /Users/ogt/awoooi/scripts/verify-sse.js)",
-      "Bash(python -m scripts.fire_test_alert --type db_connection_timeout)",
-      "Bash(npm run:*)",
-      "Bash(docker-compose down:*)",
-      "Bash(docker-compose build:*)",
-      "Bash(docker-compose up:*)",
-      "Bash(pkill -f 'next dev')",
-      "Bash(node /Users/ogt/awoooi/scripts/test-approval-flow.js)",
-      "Bash(python -m scripts.fire_test_alert --type pod_crash)",
-      "Bash(node /Users/ogt/awoooi/scripts/test-k8s-executor.js)",
-      "Bash(kubectl cluster-info:*)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl cluster-info)",
-      "Bash(ls -la /Users/ogt/awoooi/apps/web/src/app/[locale]/)",
-      "Bash(python -c \"from src.api.v1 import audit_logs; print\\(''API module loads OK''\\)\")",
-      "Bash(curl -s http://localhost:3000/zh-TW/action-logs)",
-      "Bash(pnpm build:*)",
-      "Bash(curl -s http://localhost:8000/api/v1/audit-logs)",
-      "Bash(xargs -r kill -9 2)",
-      "Bash(/dev/null source:*)",
-      "Bash(python -c \"from opentelemetry.instrumentation.httpx import HTTPXClientInstrumentor; print\\(''''httpx ok''''\\)\")",
-      "Bash(sqlite3 /Users/ogt/awoooi/apps/api/awoooi.db \"SELECT * FROM audit_logs ORDER BY created_at DESC LIMIT 5;\")",
-      "Bash(sqlite3 /Users/ogt/awoooi/apps/api/awoooi.db \"SELECT name FROM sqlite_master WHERE type=''table'';\")",
-      "Bash(sqlite3 /Users/ogt/awoooi/apps/api/awoooi.db \"SELECT id, event_type, status, title, created_at FROM timeline_events ORDER BY created_at DESC LIMIT 5;\")",
-      "Bash(curl -s http://localhost:8000/api/v1/audit-logs/stats)",
-      "Bash(curl -s http://localhost:8000/api/v1/timeline?limit=10)",
-      "Bash(curl -s \"http://localhost:8000/api/v1/timeline\")",
-      "Bash(curl -s http://localhost:8000/api/v1/docs)",
-      "Bash(chmod +x /Users/ogt/awoooi/scripts/setup-guardrails.sh /Users/ogt/awoooi/scripts/ai_code_reviewer.py)",
-      "Bash(ls -la /Users/ogt/awoooi/apps/web/.eslintrc*)",
-      "Bash(ls -la scripts/*.py scripts/*.sh .pre-commit-config.yaml .secrets.baseline apps/web/.eslintrc.js)",
-      "Bash(python -m src.services.test_context_gatherer)",
-      "Bash(python -m pytest src/services/test_context_gatherer.py -v)",
-      "Bash(grep -r \"ClawBot\\\\|clawbot\\\\|CLAWBOT\" --include=*.py --include=*.ts --include=*.tsx apps/)",
-      "Bash(python scripts/e2e_openclaw_test.py)",
-      "Bash(python -m pytest tests/e2e_network_test.py -v --tb=short)",
-      "Bash(chmod +x /Users/ogt/awoooi/apps/api/scripts/apply_prometheus_config.sh /Users/ogt/awoooi/apps/api/scripts/fire_live_alert.py)",
-      "Bash(./scripts/apply_prometheus_config.sh)",
-      "Bash(python scripts/fire_live_alert.py oomkilled)",
-      "Bash(python scripts/fire_live_alert.py oomkilled --api-url http://localhost:8000)",
-      "Bash(python scripts/fire_live_alert.py highcpu --api-url http://localhost:8000)",
-      "Bash(python scripts/fire_live_alert.py podcrash --api-url http://localhost:8000)",
-      "Bash(python -m pytest tests/test_webhook_telegram_integration.py -v)",
-      "Bash(ls -la /Users/ogt/awoooi/apps/api/.env*)",
-      "Bash(ls -la /Users/ogt/wooo-aiops/.env*)",
-      "Bash(ls -la /Users/ogt/AIOps/.env*)",
-      "Bash(/Users/ogt/awoooi/apps/api/.env:*)",
-      "Bash(/tmp/deploy-188-home.sh:*)",
-      "Bash(chmod +x /tmp/deploy-188-home.sh)",
-      "Bash(scp /tmp/awoooi-api-deploy.tar.gz /tmp/deploy-188-home.sh ollama@192.168.0.188:/tmp/)",
-      "Bash(ssh ollama@192.168.0.188 \"bash /tmp/deploy-188-home.sh\")",
-      "Bash(ssh ollama@192.168.0.188 \"curl -s http://localhost:8000/api/v1/webhooks/health\")",
-      "Bash(ssh ollama@192.168.0.188 \"tail -50 /tmp/openclaw.log\")",
-      "Bash(ssh ollama@192.168.0.188 \"cd /home/ollama/awoooi-api && source .venv/bin/activate && pip install sqlalchemy aiosqlite -q && pip install httpx python-dotenv pydantic-settings -q\")",
-      "Bash(ssh ollama@192.168.0.188 \"cd /home/ollama/awoooi-api && pkill -f ''uvicorn src.main:app'' 2>/dev/null; sleep 1; source .venv/bin/activate && nohup uvicorn src.main:app --host 0.0.0.0 --port 8000 > /tmp/openclaw.log 2>&1 & sleep 3 && curl -s http://localhost:8000/api/v1/webhooks/health\")",
-      "Bash(ssh ollama@192.168.0.188:*)",
-      "Bash(pkill -f ngrok)",
-      "Bash(pkill -f \"ssh -fN.*8001\")",
-      "Bash(ssh -fN -L 8001:localhost:8000 ollama@192.168.0.188)",
-      "Bash(curl -s http://localhost:8001/api/v1/webhooks/health)",
-      "Bash(BOT_TOKEN=\"8569720657:AAHdvKf_P2ms-QKFTyqTLtLiqEggz8cpjMk\" curl -s \"https://api.telegram.org/bot$BOT_TOKEN/getWebhookInfo\")",
-      "Bash(curl -s https://api.telegram.org/bot$BOT_TOKEN/getWebhookInfo)",
-      "Bash(curl -s http://localhost:8001/api/v1/webhooks/)",
-      "Bash(curl -s http://localhost:8001/)",
-      "Bash(curl -s http://localhost:8001/api/v1/health)",
-      "Bash(scp /tmp/awoooi-api-v7.tar.gz ollama@192.168.0.188:/tmp/)",
-      "Bash(tar -czvf /tmp/awoooi-api-v7.1.tar.gz src/ requirements.txt pyproject.toml)",
-      "Bash(scp /tmp/awoooi-api-v7.1.tar.gz ollama@192.168.0.188:/tmp/)",
-      "Bash(ssh ollama@192.168.0.188 \"tail -10 /tmp/openclaw.log | grep -E ''''clickhouse|signoz_gold''''\")",
-      "Bash(ssh ogt@192.168.0.188 \"cd /home/ollama/awoooi-api && tail -50 nohup.out 2>/dev/null || journalctl -u awoooi-api --no-pager -n 50 2>/dev/null || echo ''請手動檢查日誌''\")",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8123/ -d \"SELECT 1 FORMAT JSONEachRow\")",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:11434/api/tags)",
-      "Bash(ssh -o StrictHostKeyChecking=no -o PasswordAuthentication=no -o BatchMode=yes -o ConnectTimeout=5 ollama@192.168.0.188 \"echo ok\")",
-      "Bash(ssh -o StrictHostKeyChecking=no -o PasswordAuthentication=no -o BatchMode=yes -o ConnectTimeout=5 wooo@192.168.0.188 \"echo ok\")",
-      "Bash(ssh -o StrictHostKeyChecking=no -o PasswordAuthentication=no -o BatchMode=yes -o ConnectTimeout=5 root@192.168.0.188 \"echo ok\")",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8001/health)",
-      "Bash(ssh root@192.168.0.188 \"cat /tmp/openclaw.log 2>/dev/null | tail -100 || echo ''Log file not found''\")",
-      "Bash(ssh -o StrictHostKeyChecking=no -o BatchMode=yes -o ConnectTimeout=5 ollama@192.168.0.188 \"echo ok\")",
-      "Bash(ssh -o StrictHostKeyChecking=no -o BatchMode=yes -o ConnectTimeout=5 wooo@192.168.0.188 \"echo ok\")",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/services/signoz_client.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/services/)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/services/openclaw.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/services/)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/services/telegram_gateway.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/services/)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/api/v1/webhooks.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/api/v1/)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/models/ai.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/models/)",
-      "Bash(ssh ollama@192.168.0.188 \"cd /home/ollama/awoooi-api && pkill -f ''''uvicorn src.main:app'''' && sleep 2 && nohup .venv/bin/python3 -m uvicorn src.main:app --host 0.0.0.0 --port 8000 > nohup.out 2>&1 &\")",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8000/health)",
-      "Bash(curl -s --connect-timeout 10 http://192.168.0.188:8000/health)",
-      "Bash(curl -s -X POST http://192.168.0.188:8000/api/v1/webhooks/alerts -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(curl -s -X POST http://192.168.0.188:8000/api/v1/webhooks/alerts -H \"Content-Type: application/json\" -d '{\"\"alert_type\"\":\"\"high_cpu\"\",\"\"severity\"\":\"\"critical\"\",\"\"source\"\":\"\"signoz\"\",\"\"target_resource\"\":\"\"api-gateway\"\",\"\"namespace\"\":\"\"awoooi-prod\"\",\"\"message\"\":\"\"CPU 92% test\"\"}')",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8000/api/v1/webhooks/alerts -X POST -H \"Content-Type: application/json\" -d '{\"\"alert_type\"\":\"\"high_cpu\"\",\"\"severity\"\":\"\"critical\"\",\"\"source\"\":\"\"signoz\"\",\"\"target_resource\"\":\"\"api-gateway\"\",\"\"namespace\"\":\"\"awoooi-prod\"\",\"\"message\"\":\"\"CPU 92% - 統帥全自主驗收 v2\"\"}')",
-      "Bash(curl -s --connect-timeout 30 --max-time 120 -X POST http://192.168.0.188:8000/api/v1/webhooks/alerts -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(curl -s --connect-timeout 30 --max-time 180 -X POST http://192.168.0.188:8000/api/v1/webhooks/alerts -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/webhooks/alerts -X POST -H \"Content-Type: application/json\" -d '{\"\"alert_type\"\":\"\"k8s_pod_crash\"\",\"\"severity\"\":\"\"critical\"\",\"\"source\"\":\"\"signoz\"\",\"\"target_resource\"\":\"\"inventory-api\"\",\"\"namespace\"\":\"\"commerce\"\",\"\"message\"\":\"\"Pod crash - 統帥終極驗收\"\"}' --connect-timeout 30 --max-time 180)",
-      "Bash(ssh -o ConnectTimeout=10 ollama@192.168.0.188 \"echo OK && ps aux | grep uvicorn | grep -v grep | head -2\")",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/webhooks/alerts -X POST -H \"Content-Type: application/json\" -d '{\"\"alert_type\"\":\"\"ssl_expiry\"\",\"\"severity\"\":\"\"critical\"\",\"\"source\"\":\"\"signoz\"\",\"\"target_resource\"\":\"\"nginx-ingress\"\",\"\"namespace\"\":\"\"ingress\"\",\"\"message\"\":\"\"SSL 即將過期 - 終極驗收\"\"}' --connect-timeout 30 --max-time 180)",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/webhooks/alerts -X POST -H \"Content-Type: application/json\" -d '{\"\"alert_type\"\":\"\"db_connection_timeout\"\",\"\"severity\"\":\"\"critical\"\",\"\"source\"\":\"\"signoz\"\",\"\"target_resource\"\":\"\"postgres-primary\"\",\"\"namespace\"\":\"\"database\"\",\"\"message\"\":\"\"DB 連線逾時 - SignOz 整合終極測試\"\"}' --connect-timeout 30 --max-time 180)",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/webhooks/alerts -X POST -H \"Content-Type: application/json\" -d '{\"\"alert_type\"\":\"\"service_404\"\",\"\"severity\"\":\"\"critical\"\",\"\"source\"\":\"\"signoz\"\",\"\"target_resource\"\":\"\"auth-service\"\",\"\"namespace\"\":\"\"identity\"\",\"\"message\"\":\"\"Service 404 - SignOz + Ollama 整合終極測試\"\"}' --connect-timeout 30 --max-time 180)",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/webhooks/alerts -X POST -H \"Content-Type: application/json\" -d '{\"\"alert_type\"\":\"\"high_cpu\"\",\"\"severity\"\":\"\"warning\"\",\"\"source\"\":\"\"signoz\"\",\"\"target_resource\"\":\"\"recommendation-engine\"\",\"\"namespace\"\":\"\"ml\"\",\"\"message\"\":\"\"CPU 78% - Ollama 最終測試\"\"}' --connect-timeout 30 --max-time 200)",
-      "Bash(scp apps/api/src/services/openclaw.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/services/openclaw.py)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/core/http_client.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/core/)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/main.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/core/config.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/core/)",
-      "Bash(scp /Users/ogt/awoooi/apps/api/src/api/v1/health.py ollama@192.168.0.188:/home/ollama/awoooi-api/src/api/v1/)",
-      "Bash(ssh -o ConnectTimeout=5 ollama@192.168.0.188 \"ps aux | grep uvicorn | grep -v grep\")",
-      "Bash(curl -s -H \"Origin: http://localhost:3000\" -H \"Access-Control-Request-Method: GET\" -X OPTIONS http://192.168.0.188:8000/api/v1/health -v)",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/health)",
-      "Bash(curl -s -N --max-time 3 http://192.168.0.188:8000/api/v1/dashboard/stream)",
-      "Bash(curl -s http://localhost:3000/zh-TW -o /dev/null -w \"%{http_code}\")",
-      "Bash(open http://localhost:3000/zh-TW)",
-      "Bash(open http://localhost:3001/zh-TW)",
-      "Bash(curl -s -H \"Origin: http://localhost:3001\" http://192.168.0.188:8000/api/v1/dashboard/stream --max-time 3)",
-      "Bash(curl -s -I -H \"Origin: http://localhost:3001\" http://192.168.0.188:8000/api/v1/health)",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/approvals/pending)",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/approvals)",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/approvals?status=pending_approval\")",
-      "Bash(xargs sed:*)",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/approvals/history?limit=5\")",
-      "Bash(curl -s http://192.168.0.188:8000/api/v1/approvals/approved)",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/timeline?limit=10\")",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/action-logs\")",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/timeline/events?limit=10\")",
-      "Bash(ssh ogt@192.168.0.188 \"kubectl get nodes\")",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/approvals/k8s-test\")",
-      "Bash(scp /Users/ogt/awoooi/apps/api/k3s-prod.yaml ogt@192.168.0.188:~/awoooi-api/k3s-prod.yaml)",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/timeline/events?limit=5\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.120 \"cat /etc/rancher/k3s/k3s.yaml\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no wooo@192.168.0.188 \"echo ''SSH OK'' && pwd\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"echo ''SSH OK'' && pwd && ls -la ~/awoooi-api/ 2>/dev/null || echo ''Directory not found''\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"sshpass -p ''0936223270'' scp -o StrictHostKeyChecking=no wooo@192.168.0.120:/etc/rancher/k3s/k3s.yaml ~/awoooi-api/k3s-prod.yaml && sed -i ''s/127.0.0.1/192.168.0.120/g'' ~/awoooi-api/k3s-prod.yaml && echo ''Kubeconfig deployed!'' && head -10 ~/awoooi-api/k3s-prod.yaml\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd ~/awoooi-api && pkill -f ''uvicorn'' 2>/dev/null; sleep 1; nohup .venv/bin/uvicorn src.main:app --host 0.0.0.0 --port 8000 --reload > nohup.out 2>&1 & sleep 3; echo ''=== API Restarted ==='' && tail -20 nohup.out\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 \"cd ~/awoooi-api && pkill -f ''uvicorn src.main'' || true\")",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/health\" --connect-timeout 5)",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 ollama@192.168.0.188 \"cd ~/awoooi-api && source .venv/bin/activate && nohup uvicorn src.main:app --host 0.0.0.0 --port 8000 > nohup.out 2>&1 &\")",
-      "Bash(sshpass -p:*)",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/health\" --connect-timeout 10)",
-      "Bash(curl -s \"http://192.168.0.188:8000/api/v1/timeline/events?limit=8\")",
-      "Bash(curl -s http://localhost:3000/zh-TW -o /dev/null -w \"Frontend: HTTP %{http_code}\\\\n\")",
-      "Bash(sshpass -p '0936223270' ssh -o StrictHostKeyChecking=no ollama@192.168.0.188 'curl -s http://localhost:8000/api/v1/approvals/pending | jq -r \"\".approvals[] | \\\\\"\"ID: \\\\\\(.id\\) | Action: \\\\\\(.action\\)\\\\\"\"\"\"')",
-      "Bash(curl -s --connect-timeout 5 https://awoooi.wooo.tw/api/v1/health)",
-      "Bash(curl -s --connect-timeout 5 https://awoooi.wooo.tw/api/v1/approvals/pending)",
-      "Bash(ssh ollama@192.168.70.188 \"ps aux | grep uvicorn | grep -v grep | head -3\")",
-      "Bash(ssh -o ConnectTimeout=10 ollama@192.168.70.188 \"echo ''SSH Connected''\")",
-      "Bash(ping -c 2 -t 5 192.168.70.188)",
-      "Bash(curl -s --connect-timeout 10 https://awoooi.wooo.tw/api/v1/health)",
-      "Bash(ssh -o ConnectTimeout=10 ollama@192.168.0.188 \"echo ''SSH Connected to 188 Base''\")",
-      "Bash(grep -B 5 -A 30 \"async def add_signature\" /Users/ogt/awoooi/apps/api/src/services/*.py)",
-      "Bash(ssh ogt@192.168.0.188 \"cd /home/ogt/awoooi && docker compose ps\")",
-      "Bash(ls -la .env*)",
-      "Bash(.env:*)",
-      "Bash(timeout 15 python -m uvicorn src.main:app --host 0.0.0.0 --port 8001)",
-      "Bash(timeout 20 python -m uvicorn src.main:app --host 0.0.0.0 --port 8001)",
-      "Bash(timeout 25 python -m uvicorn src.main:app --host 0.0.0.0 --port 8001)",
-      "Bash(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no ogt@192.168.0.188 \"cd /home/ogt/wooo-aiops && docker compose ps clawbot 2>/dev/null || docker ps | grep -i claw\")",
-      "Bash(ls -la ~/.ssh/*.pub)",
-      "Bash(ssh -i ~/.ssh/id_rsa -o ConnectTimeout=5 -o StrictHostKeyChecking=no -o PasswordAuthentication=no ogt@192.168.0.188 \"echo connected\")",
-      "Bash(curl -s \"https://api.telegram.org/bot8569720657:AAHdvKf_P2ms-QKFTyqTLtLiqEggz8cpjMk/logOut\")",
-      "Bash(curl -s \"https://api.telegram.org/bot8569720657:AAHdvKf_P2ms-QKFTyqTLtLiqEggz8cpjMk/close\")",
-      "Bash(curl -s \"https://api.telegram.org/bot8569720657:AAHdvKf_P2ms-QKFTyqTLtLiqEggz8cpjMk/getUpdates?timeout=3&limit=1\")",
-      "Bash(ping -c 1 192.168.0.188)",
-      "Bash(python -m tests.test_redis_multisig)",
-      "Bash(curl -v -X POST http://localhost:8000/api/v1/webhooks/signals -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(python3 -c \":*)",
-      "Bash(echo '  無法連線' __NEW_LINE_8fc87454f9798a7d__ echo echo [結論]: echo '  /signals 端點尚未部署到 .188' echo '  程式碼已完成，需要執行:' echo \"    cd apps/api && docker build -t awoooi-api . && docker-compose up -d\")",
-      "Bash(__NEW_LINE_dc88f37970737861__ cd:*)",
-      "Bash(__NEW_LINE_dc88f37970737861__ echo:*)",
-      "Read(//Users/**)",
-      "Bash(tail -20 __NEW_LINE_8b049957a9782734__ echo \"\" echo \"[Step 2] 等待容器啟動 \\(10 秒\\)...\" sleep 10 __NEW_LINE_8b049957a9782734__ echo \"\" echo \"[Step 3] 檢查容器狀態...\" docker compose ps)",
-      "Bash(tail -5 __NEW_LINE_275e0094e9dcb44a__ echo \"\" echo \"[1.2] 重建 API 容器 \\(含 Signal Worker\\)...\" docker compose build api)",
-      "Bash(1 __NEW_LINE_275e0094e9dcb44a__ echo \"\" echo \"[1.4] 等待服務就緒 \\(15 秒\\)...\" sleep 15 __NEW_LINE_275e0094e9dcb44a__ echo \"\" echo \"[1.5] 檢查容器狀態...\" docker compose ps)",
-      "Bash(__NEW_LINE_f4c8301ec5249760__ echo:*)",
-      "Bash(__NEW_LINE_21ba3cf3700d942d__ cd:*)",
-      "Bash(1 __NEW_LINE_9a14b79fc58c11ba__ echo \"\" echo \"[1.3] 等待服務就緒 \\(15 秒\\)...\" sleep 15 __NEW_LINE_9a14b79fc58c11ba__ echo \"\" echo \"[1.4] 檢查容器狀態...\" docker compose ps api)",
-      "Bash(1 __NEW_LINE_6b654ca5be87c137__ echo \"\" echo \"[2] 等待服務就緒 \\(15 秒\\)...\" sleep 15 __NEW_LINE_6b654ca5be87c137__ echo \"\" echo \"[3] 發送測試 Signal...\" curl -s -X POST http://localhost:8000/api/v1/webhooks/signals -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(__NEW_LINE_564908ddf866c081__ echo:*)",
-      "Bash(chmod +x /Users/ogt/awoooi/apps/api/scripts/test_phase63_aggregation.py)",
-      "Bash(python scripts/test_phase63_aggregation.py)",
-      "Bash(xargs -r docker exec -i awoooi-redis redis-cli DEL)",
-      "Bash(chmod +x /Users/ogt/awoooi/apps/api/scripts/test_race_condition.py)",
-      "Bash(python scripts/test_race_condition.py)",
-      "Bash(chmod +x /Users/ogt/awoooi/apps/api/scripts/test_phase64_proposal.py)",
-      "Bash(python scripts/test_phase64_proposal.py)",
-      "Bash(python agent.py --alert FINAL_PHASE_6_TEST)",
-      "Bash(AWOOOI_REDIS_URL=\"redis://localhost:6379/0\" python agent.py --alert FINAL_PHASE_6_TEST)",
-      "Bash(curl -s http://localhost:8000/api/v1/incidents)",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/incidents/INC-20260322-06085B/proposal)",
-      "Bash(grep -r \"mock\\\\|Mock\\\\|MOCK\\\\|fake\\\\|Fake\\\\|dummy\\\\|hardcode\" /Users/ogt/awoooi/apps/web/src --include=*.tsx --include=*.ts -l)",
-      "Bash(NEXT_PUBLIC_API_URL=http://localhost:8000 pnpm next build --no-lint)",
-      "Bash(grep -v \"Traceback\\\\|File \"\"/usr\\\\|^\\\\s*$\")",
-      "Bash(python -c \"import sys,json; d=json.load\\(sys.stdin\\); print\\(f''''Signal Count: {len\\(d[\"\"signals\"\"]\\)}''''\\); [print\\(f''''  - {s[\"\"alert_name\"\"]} \\({s[\"\"signal_id\"\"]}\\)''''\\) for s in d[''''signals'''']]\")",
-      "Bash(curl -s -o /dev/null -w \"%{http_code}\" http://localhost:3003/zh-TW)",
-      "Bash(curl -s -X GET \"http://localhost:8000/api/v1/incidents\" -H \"Origin: http://localhost:3003\" -H \"Access-Control-Request-Method: GET\" -v)",
-      "Bash(grep -r TELEGRAM /Users/ogt/awoooi/apps/api/.env*)",
-      "Bash(grep -r TELEGRAM_BOT_TOKEN /Users/ogt/awoooi --include=*.env* --include=*.yaml --include=*.yml)",
-      "Bash(curl -s -I -X OPTIONS \"http://localhost:8000/api/v1/incidents\" -H \"Origin: http://localhost:3000\" -H \"Access-Control-Request-Method: GET\")",
-      "Bash(curl -s \"http://localhost:8000/api/v1/incidents\" -H \"Origin: http://localhost:3000\")",
-      "Bash(python /tmp/e2e_drill.py)",
-      "Bash(python -c \"import sys,json; d=json.load\\(sys.stdin\\); i=[x for x in d[''''incidents''''] if x[''''incident_id'''']==''''INC-20260322-06085B''''][0]; print\\(f\"\"Incident: {i[''''incident_id'''']}\"\"\\); print\\(f\"\"Signals: {i[''''signal_count'''']}\"\"\\); print\\(f\"\"Updated: {i[''''updated_at'''']}\"\"\\)\")",
-      "Bash(curl -s -X POST \"http://localhost:8000/api/v1/telegram/test\")",
-      "Bash(curl -s -X POST \"http://localhost:8000/api/v1/telegram/test-push\" -H \"Content-Type: application/json\" -d '{\"\"\"\"approval_id\"\"\"\": \"\"\"\"15ab6844-ca4e-4a13-aead-dc71cd342445\"\"\"\", \"\"\"\"risk_level\"\"\"\": \"\"\"\"critical\"\"\"\", \"\"\"\"resource_name\"\"\"\": \"\"\"\"api-gateway\"\"\"\", \"\"\"\"root_cause\"\"\"\": \"\"\"\"E2E DRILL - PodCrashLoopBackOff\"\"\"\", \"\"\"\"suggested_action\"\"\"\": \"\"\"\"RESTART_DEPLOYMENT\"\"\"\", \"\"\"\"estimated_downtime\"\"\"\": \"\"\"\"5-15 min\"\"\"\"}')",
-      "Bash(curl -s -o /dev/null -w \"HTTP Status: %{http_code}\\\\n\" http://localhost:3000/zh-TW)",
-      "Bash(curl -s -I \"http://localhost:8000/api/v1/incidents\" -H \"Origin: http://localhost:3000\")",
-      "Bash(curl -s -X POST http://localhost:8000/api/v1/incidents/INC-20260322-19DF60/proposal)",
-      "Bash(curl -s -X POST \"http://localhost:8000/api/v1/telegram/test-push\" -H \"Content-Type: application/json\" -d '{\"\"\"\"approval_id\"\"\"\": \"\"\"\"942e762e-fb97-480f-b21a-d3be67fa70b1\"\"\"\", \"\"\"\"risk_level\"\"\"\": \"\"\"\"critical\"\"\"\", \"\"\"\"resource_name\"\"\"\": \"\"\"\"core-system\"\"\"\", \"\"\"\"root_cause\"\"\"\": \"\"\"\"E2E DRILL TAKE 2 - 二次實彈演習\"\"\"\", \"\"\"\"suggested_action\"\"\"\": \"\"\"\"INVESTIGATE_SERVICE\"\"\"\", \"\"\"\"estimated_downtime\"\"\"\": \"\"\"\"5-15 min\"\"\"\"}')",
-      "Bash(curl -s \"http://localhost:8000/api/v1/incidents\" -H \"Origin: http://localhost:3000\" -H \"Accept: application/json\")",
-      "Bash(python -c \"import sys,json; d=json.load\\(sys.stdin\\); print\\(f''''Incidents: {d[\"\"count\"\"]}''''\\); [print\\(f''''  - {i[\"\"incident_id\"\"]} | {i[\"\"severity\"\"]} | {i[\"\"signal_count\"\"]} signals | {i[\"\"affected_services\"\"]}''''\\) for i in d[''''incidents'''']]\")",
-      "Bash(curl -s \"http://localhost:8000/api/v1/approvals/pending\" -H \"Origin: http://localhost:3000\")",
-      "Bash(python -c \"import sys,json; d=json.load\\(sys.stdin\\); print\\(f''''Pending: {d[\"\"count\"\"]} approvals''''\\); [print\\(f''''  - {a[\"\"id\"\"][:8]}... | {a[\"\"risk_level\"\"]} | {a[\"\"action\"\"][:30]}...''''\\) for a in d[''''approvals''''][:3]]\")",
-      "Bash(mkdir -p /Users/ogt/awoooi/apps/web/public/fonts)",
-      "Bash(curl -sL -o DSEG7Classic-Bold.woff2 \"https://cdn.jsdelivr.net/npm/dseg@0.46.0/fonts/DSEG7-Classic/DSEG7Classic-Bold.woff2\")",
-      "Bash(curl -sL -o DSEG7Classic-Bold.woff \"https://cdn.jsdelivr.net/npm/dseg@0.46.0/fonts/DSEG7-Classic/DSEG7Classic-Bold.woff\")",
-      "Bash(curl -sL -o DSEG7Classic-Regular.woff2 \"https://cdn.jsdelivr.net/npm/dseg@0.46.0/fonts/DSEG7-Classic/DSEG7Classic-Regular.woff2\")",
-      "Bash(curl -sL -o DSEG7Classic-Regular.woff \"https://cdn.jsdelivr.net/npm/dseg@0.46.0/fonts/DSEG7-Classic/DSEG7Classic-Regular.woff\")",
-      "Bash(pnpm next:*)",
-      "Bash(chmod +x /Users/ogt/awoooi/scripts/bootstrap_prod.sh)",
-      "Bash(/Users/ogt/awoooi/.env:*)",
-      "Bash(grep -E \"^\\\\.env$|03-secrets\\\\.yaml\" .gitignore)",
-      "Bash(echo 'Adding to .gitignore...' if ! grep -q ^.env$ .gitignore)",
-      "Bash(then echo:*)",
-      "Bash(git add:*)",
-      "Bash(git commit:*)",
-      "Bash(git push:*)",
-      "Bash(git remote:*)",
-      "Bash(gh repo:*)",
-      "Bash(gh api:*)",
-      "Bash(gh run:*)",
-      "Bash(ls -la pnpm-*.yaml package.json turbo.json)",
-      "Bash(git status:*)",
-      "Bash(gh workflow:*)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod -o wide\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-77545758fc-xnncc -n awoooi-prod --tail=50\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-77545758fc-xnncc -n awoooi-prod 2>&1 | grep -i ''cors'' -A 5 -B 5\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-79948cbbbf-b8cgj -n awoooi-prod --tail=100\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod -l app=awoooi-api --sort-by=.metadata.creationTimestamp -o name | tail -1 | xargs kubectl logs -n awoooi-prod --tail=50\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath=''{.data.OPENCLAW_TG_USER_WHITELIST}'' | base64 -d\")",
-      "Bash(ssh wooo@192.168.0.120 'kubectl patch secret awoooi-secrets -n awoooi-prod --type='\"''\"'json'\"''\"' -p='\"''\"'[:*)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout restart deployment/awoooi-api -n awoooi-prod && kubectl rollout status deployment/awoooi-api -n awoooi-prod --timeout=120s\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout restart deployment/awoooi-worker -n awoooi-prod && kubectl rollout status deployment/awoooi-worker -n awoooi-prod --timeout=120s\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-747967b787-fcx2r -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.110 \"ps aux | grep -E ''actions-runner|Runner'' | grep -v grep\")",
-      "Bash(curl -sf http://192.168.0.120:32334/api/v1/health)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-fd795cd87-rdpgn -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.110 \"curl -sf http://192.168.0.120:32334/api/v1/health | jq .status\")",
-      "Bash(ssh wooo@192.168.0.110 \"curl -sf http://192.168.0.120:32334/api/v1/health\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sf http://localhost:32334/api/v1/health\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get svc -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sf http://10.43.125.201:8000/api/v1/health\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sf http://10.43.105.105:3000/ -o /dev/null && echo ''Web OK''\")",
-      "Bash(ssh ogt@192.168.0.188 \"ls -la /etc/nginx/sites-available/\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=50\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-795c95ff76-wch2p -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod && ss -tlnp | grep 32334\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sf http://127.0.0.1:32334/api/v1/health | head -c 200\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo ufw status 2>/dev/null || sudo iptables -L INPUT -n | head -20\")",
-      "Bash(ssh wooo@192.168.0.110 \"curl -sf --connect-timeout 5 http://192.168.0.120:32334/api/v1/health | head -c 100\")",
-      "Bash(ssh wooo@192.168.0.110 \"curl -v --connect-timeout 5 http://192.168.0.120:32334/api/v1/health 2>&1 | head -30\")",
-      "Bash(ssh wooo@192.168.0.120 \"cat /etc/systemd/system/k3s.service 2>/dev/null | grep -i exec || ps aux | grep k3s | head -3\")",
-      "Bash(ssh wooo@192.168.0.120 \"cat /etc/systemd/system/k3s.service\")",
-      "Bash(ssh wooo@192.168.0.120 \"netstat -tlnp 2>/dev/null | grep 32334 || ss -tlnp | grep 32334\")",
-      "Bash(ssh wooo@192.168.0.110 \"curl -sf --connect-timeout 5 http://192.168.0.120:31234/health 2>&1 | head -c 100\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get networkpolicy -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get networkpolicy allow-nginx-ingress -n awoooi-prod -o yaml\")",
-      "Bash(curl -sk https://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -sk -I -X OPTIONS https://awoooi.wooo.work/api/v1/health -H \"Origin: https://awoooi.wooo.work\" -H \"Access-Control-Request-Method: GET\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sI --connect-timeout 3 http://127.0.0.1:32334/api/v1/health 2>&1 | head -5\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sI --connect-timeout 3 http://127.0.0.1:32335/ 2>&1 | head -5\")",
-      "Bash(ssh wooo@192.168.0.121 \"curl -sI --connect-timeout 3 http://127.0.0.1:32334/api/v1/health 2>&1 | head -5\")",
-      "Bash(ssh wooo@192.168.0.121 \"curl -sI --connect-timeout 3 http://127.0.0.1:32335/ 2>&1 | head -5\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo iptables -t nat -L KUBE-NODEPORTS -n 2>/dev/null | head -20\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo netstat -tlnp | grep -E ''32334|32335''\")",
-      "Bash(ssh wooo@192.168.0.120 \"ss -tlnp 2>/dev/null | grep -E ''32334|32335'' || netstat -tln | grep -E ''32334|32335''\")",
-      "Bash(ssh wooo@192.168.0.120 \"ss -tln | grep -E ''32334|32335|:323''\")",
-      "Bash(ssh wooo@192.168.0.120 \"ss -tln\")",
-      "Bash(ssh wooo@192.168.0.120 \"export KUBECONFIG=/home/wooo/.kube/config-120; /home/wooo/bin/kubectl get svc -n awoooi-prod -o wide\")",
-      "Bash(ssh wooo@192.168.0.120 \"which kubectl || find /usr -name kubectl 2>/dev/null | head -1\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get svc -n awoooi-prod && kubectl get pods -n awoooi-prod -o wide\")",
-      "Bash(ssh wooo@192.168.0.120 \"export KUBECONFIG=/home/wooo/.kube/config-120 && kubectl logs awoooi-api-546b88465d-lb8zm -n awoooi-prod --tail 80\")",
-      "Bash(ssh wooo@192.168.0.120 \"KUBECONFIG=/home/wooo/.kube/config-120 kubectl logs awoooi-api-546b88465d-lb8zm -n awoooi-prod --tail 80 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"ls -la /home/wooo/.kube/ && cat /home/wooo/.kube/config-120 2>/dev/null | head -20 || cat /etc/rancher/k3s/k3s.yaml 2>/dev/null | head -20\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo cat /etc/rancher/k3s/k3s.yaml | head -20\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && kubectl logs awoooi-api-546b88465d-lb8zm -n awoooi-prod --tail 100 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"which kubectl 2>/dev/null || find /home/wooo -name kubectl 2>/dev/null | head -1 || ls -la /home/wooo/bin/\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl logs awoooi-api-546b88465d-lb8zm -n awoooi-prod --tail 100 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl describe pod awoooi-api-546b88465d-lb8zm -n awoooi-prod | tail -40\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get svc -n awoooi-prod -o wide\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl exec -n awoooi-prod deploy/awoooi-api -- curl -sf http://localhost:8000/api/v1/health 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl exec -n awoooi-prod deploy/awoooi-api -- wget -qO- http://localhost:8000/api/v1/health 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl logs deployment/awoooi-api -n awoooi-prod --tail 20 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"curl -sf http://192.168.0.120:32334/api/v1/health 2>&1 || echo ''FAILED to connect to 120:32334''\")",
-      "Bash(ssh wooo@192.168.0.110 \"curl -sf http://192.168.0.121:32334/api/v1/health 2>&1 || echo ''FAILED to connect to 121:32334''\")",
-      "Bash(ssh wooo@192.168.0.110 \"ssh wooo@192.168.0.120 ''cat /etc/rancher/k3s/k3s.yaml 2>/dev/null || echo No k3s.yaml''\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get pods -n awoooi-prod -o wide | grep Running\")",
-      "Bash(ssh -o ConnectTimeout=5 wooo@192.168.0.120 \"ufw status 2>/dev/null || firewall-cmd --state 2>/dev/null || echo ''No firewall command found''\")",
-      "Bash(ssh -o ConnectTimeout=5 wooo@192.168.0.121 \"ufw status 2>/dev/null || firewall-cmd --state 2>/dev/null || echo ''No firewall command found''\")",
-      "Bash(pip3 show:*)",
-      "Bash(docker build:*)",
-      "Bash(docker version:*)",
-      "Bash(docker run:*)",
-      "Bash(curl -vI -H \"Origin: https://awoooi.wooo.work\" http://localhost:8889/api/v1/health)",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get endpoints awoooi-api-svc -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get pods -n awoooi-prod -o wide\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo -n ufw status 2>/dev/null || sudo -n iptables -L INPUT -n 2>/dev/null | head -20 || echo ''Need sudo for firewall check''\")",
-      "Bash(ssh wooo@192.168.0.120 \"ss -tln | grep -E ''32334|32335|:323'' || echo ''No NodePort listeners found''\")",
-      "Bash(ssh wooo@192.168.0.121 \"ss -tln | grep -E ''32334|32335|:323'' || echo ''No NodePort listeners found''\")",
-      "Bash(ssh wooo@192.168.0.120 \"ps aux | grep -E ''kube-proxy|k3s'' | grep -v grep | head -5\")",
-      "Bash(ssh wooo@192.168.0.120 \"cat /proc/sys/net/ipv4/ip_forward\")",
-      "Bash(ssh wooo@192.168.0.120 \"systemctl status k3s 2>/dev/null | head -15 || ps aux | grep ''k3s server'' | grep -v grep\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sf --connect-timeout 5 http://127.0.0.1:32334/api/v1/health 2>&1 || echo ''LOCALHOST NodePort FAILED''\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -sf --connect-timeout 5 http://192.168.0.120:32334/api/v1/health 2>&1 || echo ''EXTERNAL IP NodePort FAILED''\")",
-      "Bash(ssh wooo@192.168.0.120 \"cat /etc/iptables/rules.v4 2>/dev/null || iptables-save 2>/dev/null | grep -E ''DROP|REJECT|32334|32335'' | head -10 || echo ''Cannot read iptables without sudo''\")",
-      "Bash(ssh wooo@192.168.0.121 \"curl -sf --connect-timeout 5 http://192.168.0.120:32334/api/v1/health 2>&1 || echo ''Worker->Master NodePort FAILED''\")",
-      "Bash(ssh wooo@192.168.0.120 \"cat /etc/rancher/k3s/config.yaml 2>/dev/null || ls -la /etc/rancher/k3s/ 2>/dev/null || echo ''No K3s config found''\")",
-      "Bash(ssh wooo@192.168.0.120 \"netstat -an 2>/dev/null | grep 32334 || ss -an | grep 32334 || echo ''No socket found for 32334''\")",
-      "Bash(ssh wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S iptables -L INPUT -n 2>&1 | head -20\")",
-      "Bash(ssh wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S iptables -t nat -L KUBE-NODEPORTS -n 2>&1 | head -20\")",
-      "Bash(ssh wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S iptables -L KUBE-ROUTER-INPUT -n 2>&1 | head -30\")",
-      "Bash(ssh wooo@192.168.0.120 \"echo ''0936223270'' | sudo -S iptables -t nat -L KUBE-NODEPORTS -n 2>&1 | grep -i awoooi || echo ''NO AWOOOI RULES FOUND''\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get svc awoooi-api-svc -n awoooi-prod -o yaml | grep -A5 ''spec:''\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get networkpolicy -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl apply -f - 2>&1\")",
-      "Bash(curl -sf --connect-timeout 10 https://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -skf --connect-timeout 10 https://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -sI https://awoooi.wooo.work/)",
-      "Bash(curl -skI https://awoooi.wooo.work/)",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl logs deployment/awoooi-api -n awoooi-prod --tail 50 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl rollout restart deployment/awoooi-api -n awoooi-prod && /home/wooo/kubectl rollout status deployment/awoooi-api -n awoooi-prod --timeout=120s\")",
-      "Bash(curl -sf https://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -skf https://awoooi.wooo.work/api/v1/health)",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl logs deployment/awoooi-api -n awoooi-prod --tail 40 2>&1\")",
-      "Bash(for i:*)",
-      "Bash(do curl:*)",
-      "Bash(echo \"Request $i sent\")",
-      "Bash(done)",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl logs deployment/awoooi-api -n awoooi-prod --tail 100 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl logs deployment/awoooi-api -n awoooi-prod --tail 30 2>&1\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get configmap awoooi-config -n awoooi-prod -o yaml | grep OTEL\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl exec deployment/awoooi-api -n awoooi-prod -- env | grep OTEL\")",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl exec deployment/awoooi-api -n awoooi-prod -- python -c \"\"import socket; s=socket.socket\\(\\); s.settimeout\\(5\\); s.connect\\(\\(''192.168.0.188'', 24317\\)\\); print\\(''✅ Connection to 24317 OK''\\); s.close\\(\\)\"\" 2>&1\")",
-      "Bash(curl -vI https://awoooi.wooo.work)",
-      "Bash(curl -vI https://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -sf -X POST https://awoooi.wooo.work/api/v1/webhooks/signals -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(curl -s -X POST https://awoooi.wooo.work/api/v1/webhooks/signals -H \"Content-Type: application/json\" -d '{\"\"source\"\": \"\"prometheus\"\", \"\"severity\"\": \"\"P1\"\", \"\"message\"\": \"\"Test alert from CLI\"\"}')",
-      "Bash(curl -s -X POST https://awoooi.wooo.work/api/v1/webhooks/signals -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(ssh wooo@192.168.0.110 \"export KUBECONFIG=/home/wooo/.kube/config-120 && /home/wooo/kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath=''''{.data.WEBHOOK_HMAC_SECRET}'''' 2>/dev/null\")",
-      "Bash(timeout 15 curl -N -s https://awoooi.wooo.work/api/v1/dashboard/stream)",
-      "Bash(bash:*)",
-      "Bash(curl -s https://awoooi.wooo.work/api/v1/metrics/gold)",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT DISTINCT metric_name FROM signoz_metrics.distributed_samples_v4 WHERE unix_milli > \\(toUnixTimestamp\\(now\\(\\)\\) - 1800\\) * 1000 LIMIT 20 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT count\\(\\) as trace_count FROM signoz_traces.distributed_signoz_index_v2 WHERE timestamp > now\\(\\) - INTERVAL 30 MINUTE FORMAT TabSeparated\")",
-      "Bash(ssh wooo@192.168.0.120 \"KUBECONFIG=/home/wooo/.kube/config-120 /home/wooo/bin/kubectl get configmap awoooi-config -n awoooi-prod -o jsonpath=''{.data}'' | python3 -m json.tool 2>/dev/null | head -30\")",
-      "Bash(ssh wooo@192.168.0.120 \"KUBECONFIG=/home/wooo/.kube/config-120 /home/wooo/bin/kubectl logs deployment/awoooi-api -n awoooi-prod --tail 50 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"which kubectl || ls -la ~/bin/kubectl 2>/dev/null || ls -la /usr/local/bin/kubectl 2>/dev/null || echo ''kubectl not found''\")",
-      "Bash(ssh wooo@192.168.0.120 \"export KUBECONFIG=/home/wooo/.kube/config-120 && kubectl get configmap awoooi-config -n awoooi-prod -o jsonpath=''{.data}'' 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"ls -la ~/.kube/ 2>/dev/null; cat ~/.kube/config 2>/dev/null | head -20 || echo ''checking k3s default...''; sudo cat /etc/rancher/k3s/k3s.yaml 2>/dev/null | head -5 || echo ''no k3s config''\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo k3s kubectl get configmap awoooi-config -n awoooi-prod -o yaml 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"sudo k3s kubectl logs deployment/awoooi-api -n awoooi-prod --tail 100 2>&1\")",
-      "Bash(nc -zv 192.168.0.188 24317)",
-      "Bash(curl -s http://192.168.0.188:24318/v1/traces -X POST -H \"Content-Type: application/json\" -d '{}')",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT DISTINCT serviceName, count\\(\\) as cnt FROM signoz_traces.distributed_signoz_index_v2 WHERE timestamp > now\\(\\) - INTERVAL 24 HOUR GROUP BY serviceName ORDER BY cnt DESC LIMIT 20 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"DESCRIBE TABLE signoz_traces.distributed_signoz_index_v2 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT serviceName, count\\(\\) as cnt FROM signoz_traces.distributed_signoz_index_v2 WHERE timestamp > now\\(\\) - INTERVAL 5 MINUTE GROUP BY serviceName ORDER BY cnt DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s https://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT serviceName, count\\(\\) as cnt FROM signoz_traces.distributed_signoz_index_v2 WHERE timestamp > now\\(\\) - INTERVAL 10 MINUTE GROUP BY serviceName ORDER BY cnt DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT service_name, count\\(\\) as cnt FROM signoz_logs.distributed_logs WHERE timestamp > now\\(\\) - INTERVAL 30 MINUTE GROUP BY service_name ORDER BY cnt DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SHOW TABLES FROM signoz_logs FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT count\\(\\) as total FROM signoz_logs.distributed_logs_v2 WHERE timestamp > now\\(\\) - INTERVAL 30 MINUTE FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT JSONExtractString\\(resources_string, ''service.name''\\) as svc, count\\(\\) as cnt FROM signoz_logs.distributed_logs_v2 WHERE timestamp > now\\(\\) - INTERVAL 5 MINUTE GROUP BY svc ORDER BY cnt DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"DESCRIBE TABLE signoz_logs.distributed_logs_v2 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT resources_string[''service.name''] as svc, count\\(\\) as cnt FROM signoz_logs.distributed_logs_v2 WHERE timestamp > \\(toUnixTimestamp64Nano\\(now64\\(\\)\\) - 300000000000\\) GROUP BY svc ORDER BY cnt DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT body, resources_string FROM signoz_logs.distributed_logs_v2 WHERE timestamp > \\(toUnixTimestamp64Nano\\(now64\\(\\)\\) - 60000000000\\) LIMIT 1 FORMAT JSONEachRow\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT serviceName, count\\(\\) as cnt FROM signoz_traces.distributed_signoz_index_v2 WHERE timestamp > now\\(\\) - INTERVAL 2 MINUTE GROUP BY serviceName ORDER BY cnt DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT serviceName, name, timestamp FROM signoz_traces.distributed_signoz_index_v2 WHERE timestamp > now\\(\\) - INTERVAL 5 MINUTE ORDER BY timestamp DESC LIMIT 5 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT serviceName, name, formatDateTime\\(timestamp, ''%Y-%m-%d %H:%M:%S''\\) as ts FROM signoz_traces.distributed_signoz_index_v2 ORDER BY timestamp DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT count\\(\\) FROM signoz_traces.distributed_signoz_index_v2 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT count\\(\\) FROM signoz_traces.distributed_signoz_spans FORMAT TabSeparated\")",
-      "Bash(ssh wooo@192.168.0.188 \"docker ps | grep -E ''otel|signoz''\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT metric_name, sum\\(value\\) as total FROM signoz_metrics.distributed_samples_v4 WHERE metric_name LIKE ''otelcol%span%'' AND unix_milli > \\(toUnixTimestamp\\(now\\(\\)\\) - 300\\) * 1000 GROUP BY metric_name FORMAT TabSeparated\")",
-      "Bash(for t:*)",
-      "Bash(do)",
-      "Bash(echo -n \"$t: \")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT count\\(\\) FROM signoz_traces.$t FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"SELECT serviceName, count\\(\\) as cnt FROM signoz_traces.distributed_signoz_index_v3 WHERE timestamp > now\\(\\) - INTERVAL 10 MINUTE GROUP BY serviceName ORDER BY cnt DESC LIMIT 10 FORMAT TabSeparated\")",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \":*)",
-      "Bash(curl -s 'http://192.168.0.188:8123/' --data \"DESCRIBE TABLE signoz_traces.distributed_signoz_index_v3 FORMAT TabSeparated\")",
-      "Bash(AWOOOI_API_URL=https://awoooi.wooo.work WEBHOOK_HMAC_SECRET=\"CHANGE_ME_TO_RANDOM_64_CHARS\" python scripts/fire_live_alert.py oomkilled)",
-      "Bash(timeout 10 curl -sN https://awoooi.wooo.work/api/v1/dashboard/stream)",
-      "Bash(curl -s https://awoooi.wooo.work/api/v1/dashboard)",
-      "Bash(npm list:*)",
-      "Bash(node scripts/verify-frontend.js)",
-      "Bash(node /Users/ogt/awoooi/scripts/verify-frontend.js)",
-      "Bash(python -c \"from src.services.proposal_service import ProposalService; print\\(''''✅ ProposalService OK''''\\)\")",
-      "Bash(python -c \"from src.services.openclaw import OpenClawService; print\\(''''✅ OpenClawService OK''''\\)\")",
-      "Bash(curl -s http://192.168.0.120:32334/api/v1/incidents)",
-      "Bash(jq -r \".incidents[:2] | .[] | \"\"\\\\\\(.incident_id\\) - \\\\\\(.status\\) - \\\\\\(.severity\\)\"\"\")",
-      "Bash(curl -s -X POST \"http://192.168.0.120:32334/api/v1/incidents/INC-20260322-4B3152/propose\" -H \"Content-Type: application/json\")",
-      "Bash(kubectl logs:*)",
-      "Bash(ssh ogt@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail 30\")",
-      "Bash(curl -sv -X POST \"http://192.168.0.120:32334/api/v1/incidents/INC-20260322-4B3152/propose\" -H \"Content-Type: application/json\")",
-      "Bash(curl -s http://192.168.0.120:32334/api/v1/health)",
-      "Bash(curl -s \"http://192.168.0.120:32334/api/v1/incidents/INC-20260322-4B3152\")",
-      "Bash(curl -sv \"http://192.168.0.120:32334/api/v1/incidents\")",
-      "Bash(curl -s --retry 3 --retry-delay 2 \"http://192.168.0.120:32334/api/v1/health\")",
-      "Bash(curl -s --retry 3 --retry-delay 2 http://192.168.0.120:32334/api/v1/health)",
-      "Bash(do echo:*)",
-      "Bash(curl -s -X POST \"https://awoooi.wooo.work/api/v1/incidents/INC-20260322-4B3152/propose\" -H \"Content-Type: application/json\")",
-      "Bash(curl -s -X POST \"https://awoooi.wooo.work/api/v1/incidents/INC-20260322-4B3152/proposal\" -H \"Content-Type: application/json\")",
-      "Bash(curl -s -X POST \"https://awoooi.wooo.work/api/v1/incidents/INC-20260322-D6C6A0/proposal\" -H \"Content-Type: application/json\")",
-      "Bash(curl -s http://192.168.0.120:32334/api/v1/approvals/pending)",
-      "Bash(kubectl get:*)",
-      "Bash(curl -s -w \"\\\\nHTTP_CODE: %{http_code}\\\\n\" http://192.168.0.120:32334/api/v1/health)",
-      "Bash(curl -s http://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -s http://awoooi.wooo.work/api/v1/approvals/pending)",
-      "Bash(curl -sL https://awoooi.wooo.work/api/v1/approvals/pending -k)",
-      "Bash(ssh root@192.168.0.120 \"kubectl get pods -n awoooi-prod -o wide\")",
-      "Bash(ssh root@192.168.0.120 \"kubectl logs -n awoooi-prod -l app=awoooi-api --tail=30\")",
-      "Bash(curl -sL https://awoooi.wooo.work/api/v1/timeline -k)",
-      "Bash(curl -sL https://awoooi.wooo.work/api/v1/incidents -k)",
-      "Bash(curl -sL \"https://awoooi.wooo.work/api/v1/approvals?include_history=true\" -k)",
-      "Bash(curl -sL \"https://awoooi.wooo.work/api/v1/incidents/INC-20260322-4B3152\" -k)",
-      "Bash(curl -sL \"https://awoooi.wooo.work/api/v1/audit-logs?limit=10\" -k)",
-      "Bash(curl -sL https://awoooi.wooo.work/api/v1/audit-logs?limit=10 -k)",
-      "Bash(ssh ogt@192.168.0.120 \"kubectl logs -n awoooi-prod -l app=awoooi-api --tail=100\")",
-      "Bash(ssh ogt@192.168.0.120 \"kubectl logs -n awoooi-prod -l app=awoooi-web --tail=50\")",
-      "Bash(ssh ogt@192.168.0.188 \"kubectl --kubeconfig=/etc/rancher/k3s/k3s.yaml logs -n awoooi-prod -l app=awoooi-api --tail=100 2>/dev/null || docker logs awoooi-api --tail=100 2>/dev/null\")",
-      "Bash(curl -sL \"https://awoooi.wooo.work/api/v1/approvals/pending\" -k -w \"\\\\n\\\\nHTTP: %{http_code}\\\\nTime: %{time_total}s\\\\n\")",
-      "Bash(curl -sL -X POST https://awoooi.wooo.work/api/v1/approvals/182e07c1-118a-49d7-b71c-7d33c5484d9b/sign -H 'Content-Type: application/json' -d '{\"\"\"\"signer_id\"\"\"\": \"\"\"\"test-debug\"\"\"\", \"\"\"\"signer_name\"\"\"\": \"\"\"\"Debug Test\"\"\"\", \"\"\"\"comment\"\"\"\": \"\"\"\"Testing\"\"\"\"}' -k)",
-      "Bash(curl -s https://wwooo.aiops.tw/api/v1/health)",
-      "Bash(curl -s https://wwooo.aiops.tw/api/v1/incidents?limit=5)",
-      "Bash(curl -s https://wwooo.aiops.tw/api/v1/approvals/pending)",
-      "Bash(curl -v -s \"https://wwooo.aiops.tw/api/v1/health\")",
-      "Bash(curl -s \"https://wwooo.aiops.tw/\")",
-      "Bash(curl -s --connect-timeout 5 \"http://192.168.0.120:32334/api/v1/health\")",
-      "Bash(curl -s --connect-timeout 5 \"http://192.168.0.120:32334/api/v1/incidents?limit=5\")",
-      "Bash(ssh -o ConnectTimeout=5 wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-867f67f55d-kvdl2 -n awoooi-prod --tail=50\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod | grep -E ''NAME|worker''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod | grep worker\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-5bdc5699bb-kcv9q -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get networkpolicy -n awoooi-prod -o wide\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod --show-labels | grep worker\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get networkpolicy allow-required-egress -n awoooi-prod -o yaml\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl patch networkpolicy allow-required-egress -n awoooi-prod --type=''json'' -p=''[{\"\"op\"\": \"\"replace\"\", \"\"path\"\": \"\"/spec/podSelector/matchLabels\"\", \"\"value\"\": {\"\"system\"\": \"\"awoooi\"\"}}]''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout restart deployment/awoooi-worker -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-5bdc5699bb-kcv9q -n awoooi-prod --tail=15\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=40\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod 2>&1 | grep -E ''signal_worker|redis_pool|INFO'' | tail -10\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s http://localhost:32334/api/v1/health\")",
-      "Bash(ssh wooo@192.168.0.120 'curl -s -X POST \"\"http://localhost:32334/api/v1/webhooks/signals\"\"   -H \"\"Content-Type: application/json\"\"   -d \"\"{:*)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod | grep -E ''NAME|worker|api''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod && echo ''==='' && kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s http://localhost:32334/api/v1/incidents?limit=5\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s http://localhost:32334/api/v1/approvals/pending\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod 2>&1 | head -50\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s http://localhost:32334/api/v1/health | jq ''.components''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get secret -n awoooi-prod -o name\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath=''{.data.WEBHOOK_HMAC_SECRET}'' | base64 -d\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=20 2>&1 | grep -E ''signal|incident|telegram|INFO''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s ''http://localhost:32334/api/v1/incidents?limit=5''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod 2>&1 | grep -iE ''telegram|notification|send'' | tail -10\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s ''http://localhost:32334/api/v1/approvals/pending''\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s ''http://localhost:32334/api/v1/incidents?limit=2'' && echo ''---'' && curl -s ''http://localhost:32334/api/v1/approvals/pending''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod | grep worker && echo ''---'' && kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-6b8cc94d9c-xjdwr -n awoooi-prod --tail=40\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get networkpolicy allow-required-egress -n awoooi-prod -o jsonpath=''{.spec.podSelector}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl patch networkpolicy allow-required-egress -n awoooi-prod --type=''json'' -p=''[{\"\"op\"\": \"\"replace\"\", \"\"path\"\": \"\"/spec/podSelector\"\", \"\"value\"\": {\"\"matchLabels\"\": {\"\"system\"\": \"\"awoooi\"\"}}}]''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl delete pod awoooi-worker-6b8cc94d9c-xjdwr -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-6b8cc94d9c-pmzj7 -n awoooi-prod --tail=30\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-6b8cc94d9c-pmzj7 -n awoooi-prod --tail=20\")",
-      "Bash(ls -la /Users/ogt/awoooi/apps/api/scripts/fire*.py)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=50\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s ''http://localhost:32334/api/v1/incidents?limit=3''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod 2>&1 | grep -iE ''proposal|approval|llm|ai|ollama|generate'' | tail -20\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get deployment awoooi-worker -n awoooi-prod -o jsonpath=''{.spec.template.spec.containers[0].envFrom}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get deployment awoooi-api -n awoooi-prod -o jsonpath=''{.spec.template.spec.containers[0].envFrom}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get configmap awoooi-config -n awoooi-prod -o jsonpath=''''{.data}''''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath=''{.data}'' | tr '','' ''\\\\n''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl exec deployment/awoooi-api -n awoooi-prod -- python -c ''import os; print\\(os.getenv\\(\"\"DATABASE_URL\"\", \"\"NOT SET\"\"\\)[:50]\\)''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-75ffbfb88b-2htfh -n awoooi-prod --tail=50\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl exec awoooi-api-6687db5564-rv755 -n awoooi-prod -- env | grep DATABASE\")",
-      "Bash(ssh wooo@192.168.0.120 \"PGPASSWORD=''CHANGE_ME'' psql -h 192.168.0.188 -U awoooi -d awoooi_prod -c ''SELECT 1'' 2>&1 || echo ''Connection failed''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod\")",
-      "Bash(curl -sv http://192.168.0.120:32334/api/v1/health)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-75ffbfb88b-2htfh -n awoooi-prod --tail=20 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-7fb7d5b55f-n48gk -n awoooi-prod --tail=20 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get rs -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl scale rs awoooi-api-75ffbfb88b -n awoooi-prod --replicas=0\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl scale rs awoooi-worker-7fb7d5b55f -n awoooi-prod --replicas=0\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=10\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get deploy -n awoooi-prod -o wide\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get deploy awoooi-api -n awoooi-prod -o jsonpath=''{.spec.replicas}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get deploy awoooi-worker -n awoooi-prod -o jsonpath=''{.spec.replicas}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout status deployment/awoooi-api -n awoooi-prod --timeout=5s\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout history deployment/awoooi-api -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout undo deployment/awoooi-api -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout undo deployment/awoooi-worker -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout status deployment/awoooi-api -n awoooi-prod --timeout=30s\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get rs awoooi-api-6687db5564 -n awoooi-prod -o jsonpath=''{.metadata.annotations.deployment\\\\.kubernetes\\\\.io/revision}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl delete pod awoooi-api-7f487f7cbb-5f88g -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout undo deployment/awoooi-api -n awoooi-prod --to-revision=46\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --tail=15\")",
-      "Bash(curl -s http://192.168.0.120:32334/api/v1/incidents?limit=3)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --since=2m\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --since=2m | grep -i webhook\")",
-      "Bash(curl -sv -X POST http://192.168.0.120:32334/api/v1/webhooks/alertmanager -H \"Content-Type: application/json\" -d '{:*)",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get endpoints -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"curl -s http://localhost:32334/api/v1/health | jq ''{status}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-worker -n awoooi-prod --since=30s\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-fc4744758-7wfv5 -n awoooi-prod --tail=30 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-worker-6fc548887b-b9mtf -n awoooi-prod --tail=30 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get configmap awoooi-config -n awoooi-prod -o yaml\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath=''''{.data}''''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pod awoooi-worker-6fc548887b-b9mtf -n awoooi-prod -o jsonpath=''{.metadata.labels}''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get networkpolicy -n awoooi-prod -o yaml\")",
-      "Bash(ssh wooo@192.168.0.120 'kubectl patch networkpolicy allow-required-egress -n awoooi-prod --type=json -p=\"\"[{\\\\\"\"op\\\\\"\": \\\\\"\"replace\\\\\"\", \\\\\"\"path\\\\\"\": \\\\\"\"/spec/podSelector/matchLabels\\\\\"\", \\\\\"\"value\\\\\"\": {\\\\\"\"system\\\\\"\": \\\\\"\"awoooi\\\\\"\"}}]\"\"')",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout restart deployment/awoooi-api deployment/awoooi-worker -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs awoooi-api-6c69b77894-d6jqq -n awoooi-prod --tail=20\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl run nc-test --rm -it --restart=Never --image=busybox -- nc -zv 192.168.0.188 5432\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get pods -n awoooi-prod -o=custom-columns=''NAME:.metadata.name,IMAGE:.spec.containers[0].image''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl exec awoooi-api-6687db5564-rv755 -n awoooi-prod -- ls -la *.db 2>/dev/null || echo ''No SQLite files''\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl exec awoooi-api-6687db5564-rv755 -n awoooi-prod -- env | grep -E ''MOCK|DATABASE|SQLITE''\")",
-      "Bash(curl -s \"http://192.168.0.120:32334/api/v1/approvals\")",
-      "Bash(python -m py_compile src/lewooogo_brain/engines/incident_engine.py src/lewooogo_brain/engines/proposal_engine.py src/lewooogo_brain/skills/loader.py)",
-      "Bash(python packages/lewooogo-brain/tests/test_skill_loader.py)",
-      "Bash(python packages/lewooogo-brain/tests/test_incident_engine.py)",
-      "Bash(python packages/lewooogo-brain/tests/test_guardrails.py)",
-      "Bash(python -m py_compile src/lewooogo_brain/engines/proposal_engine.py src/lewooogo_brain/engines/incident_engine.py src/lewooogo_brain/skills/loader.py)",
-      "Bash(PYTHONPATH=/Users/ogt/awoooi/packages/lewooogo-brain/src python -c \":*)",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8000/api/v1/health)",
-      "Bash(curl -s \"https://awoooi.wooo.work/api/v1/approvals/pending\")",
-      "Bash(curl -s \"https://awoooi.wooo.work/api/v1/approvals?status=pending\")",
-      "Bash(curl -s \"https://awoooi.wooo.work/api/v1/incidents\")",
-      "Bash(uv sync:*)",
-      "Bash(python -c \"from src.routers.proposals import router; print\\(''✅ Router 語法驗證通過''\\)\")",
-      "Bash(curl -s -X GET \"https://awoooi.wooo.work/api/v1/health\" --connect-timeout 10)",
-      "Bash(curl -s -X GET \"https://awoooi.wooo.work/api/v1/incidents\" --connect-timeout 10)",
-      "Bash(curl -s -o /dev/null -w \"%{http_code}\" \"https://awoooi.wooo.work\" --connect-timeout 10)",
-      "Bash(curl -s -o /dev/null -w \"%{http_code}\" -L \"https://awoooi.wooo.work\" --connect-timeout 10)",
-      "Bash(curl -s -X POST \"https://awoooi.wooo.work/api/v1/incidents/test-123/propose\" -H \"Content-Type: application/json\" -d '{\"\"require_dry_run\"\": true}' --connect-timeout 10)",
-      "Bash(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no ollama@192.168.0.120 \"kubectl get pods -n awoooi-prod -o wide\")",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get pods -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs awoooi-api-64c8659cff-grslz -n awoooi-prod --tail=50)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath='{.data.DATABASE_URL}')",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl rollout restart deployment/awoooi-api -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get pods -n awoooi-prod -l app=awoooi-api)",
-      "Bash(curl -s \"https://awoooi.wooo.work/api/v1/health\" --connect-timeout 10)",
-      "Bash(curl -s -o /dev/null -w \"%{http_code}\" -L \"https://awoooi.wooo.work/zh-TW\" --connect-timeout 10)",
-      "Bash(python -c \"from src.routers.proposals import router; print\\(''✅ Router import successful''\\)\")",
-      "Bash(PGPASSWORD=postgres psql -h 192.168.0.188 -U awoooi -d awoooi_dev -c \"SELECT incident_id, status, severity FROM incidents LIMIT 5;\")",
-      "Bash(PGPASSWORD=AwoooiProd2026 psql -h 192.168.0.188 -U awoooi -d awoooi_prod -c \"SELECT incident_id, status, severity FROM incidents LIMIT 5;\")",
-      "Bash(curl -sf http://192.168.0.120:32334/api/v1/incidents)",
-      "Bash(curl -v \"http://192.168.0.120:32334/api/v1/incidents\")",
-      "Bash(export KUBECONFIG=/Users/ogt/.kube/config-120)",
-      "Bash(curl -sI \"http://awoooi.wooo.work/\")",
-      "Bash(openssl s_client -servername awoooi.wooo.work -connect awoooi.wooo.work:443)",
-      "Bash(openssl x509:*)",
-      "Bash(curl -s -X POST \"http://192.168.0.120:32334/api/v1/incidents/INC-20260323-7DE10B/propose\" -H \"Content-Type: application/json\" -d '{\"\"\"\"require_dry_run\"\"\"\": true}')",
-      "Bash(python -c \"from src.services.executor import execute_approved_proposal, get_executor, ActionExecutor; print\\(''✅ Import successful''\\)\")",
-      "Bash(curl -s https://awoooi.woooo.cc/api/v1/incidents)",
-      "Bash(curl -s https://awoooi.woooo.cc/api/v1/health)",
-      "Bash(curl -s --connect-timeout 10 https://awoooi.woooo.cc/api/v1/health)",
-      "Bash(ssh ogt@192.168.70.202 \"sudo kubectl get pods -n awoooi 2>/dev/null\")",
-      "Bash(curl -s --connect-timeout 5 http://192.168.70.200:8000/api/v1/health)",
-      "Bash(ssh ogt@192.168.70.202 \"sudo kubectl get pods -n awoooi-prod\")",
-      "Bash(ssh -o StrictHostKeyChecking=no ogt@192.168.70.202 \"sudo kubectl get pods -n awoooi-prod\")",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get pods -A)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod awoooi-worker-7479556d76-jbbps --tail 30)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod -l app=awoooi-api --tail 20)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl exec -n awoooi-prod deployment/awoooi-api -- curl -s http://localhost:8000/api/v1/incidents)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl exec -n awoooi-prod deployment/awoooi-api -- python -c \"import httpx; r = httpx.get\\(''http://localhost:8000/api/v1/incidents''\\); print\\(r.text\\)\")",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get ingress -n awoooi-prod -o wide)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get svc -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get deployment awoooi-worker -n awoooi-prod -o jsonpath='{.spec.template.spec.containers[0].env}')",
-      "Bash(curl -s --connect-timeout 5 http://192.168.70.202:32334/api/v1/health)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl describe deployment awoooi-worker -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get configmap -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl describe deployment awoooi-api -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get configmap awoooi-config -n awoooi-prod -o yaml)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get secrets -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath='{.data}')",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get secret awoooi-secrets -n awoooi-prod -o jsonpath='{.data.REDIS_URL}')",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl rollout restart deployment/awoooi-worker -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get pods -n awoooi-prod -l app=awoooi-worker)",
-      "Bash(curl -s --connect-timeout 5 https://awoooi.wooo.work/api/v1/health)",
-      "Bash(curl -s https://awoooi.wooo.work/api/v1/incidents)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod -l app=awoooi-worker --tail 10)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get svc -n wooo-aiops-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get svc -A)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod awoooi-worker-76bdf9786d-rvtmz --tail 15)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl exec -n awoooi-prod deployment/awoooi-api -- python -c \"import os; print\\(os.getenv\\(''REDIS_URL'', ''NOT_SET''\\)\\)\")",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get deployment awoooi-api -n awoooi-prod -o yaml)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl rollout restart deployment/awoooi-api deployment/awoooi-worker -n awoooi-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod awoooi-api-865cdc97db-6mpzz --tail 20)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get pods -n wooo-aiops-prod -l app=redis)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get pods -n wooo-aiops-prod)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl exec -n wooo-aiops-prod redis-6c6fcd64b8-8wznx -- redis-cli ping)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl exec -n awoooi-prod awoooi-api-6445c76797-mrl7p -- python -c \"import redis; r=redis.Redis\\(host=''10.43.239.47'', port=6379, db=10\\); print\\(r.ping\\(\\)\\)\")",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get networkpolicy -A)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get networkpolicy allow-required-egress -n awoooi-prod -o yaml)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl patch networkpolicy allow-required-egress -n awoooi-prod --type='json' -p='[{\"\"op\"\": \"\"add\"\", \"\"path\"\": \"\"/spec/egress/0/ports/-\"\", \"\"value\"\": {\"\"port\"\": 6379, \"\"protocol\"\": \"\"TCP\"\"}}]')",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod awoooi-api-5fcc484b85-qpwt6 --tail 15)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl exec -n awoooi-prod awoooi-api-6445c76797-mrl7p -- python -c \"import os; print\\(''REDIS_URL:'', os.getenv\\(''REDIS_URL''\\)\\); import redis; r=redis.Redis.from_url\\(os.getenv\\(''REDIS_URL''\\)\\); print\\(''PING:'', r.ping\\(\\)\\)\")",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod awoooi-worker-59d7588d75-p5tht --tail 20)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod -l app=awoooi-worker --tail 30)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get deployment awoooi-worker -n awoooi-prod -o yaml)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get networkpolicy -n awoooi-prod -o wide)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl apply -f -)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs -n awoooi-prod awoooi-worker-6cd7dcbc9-5mtfq --tail 15)",
-      "Bash(jq .incidents[0])",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl get configmap awoooi-config -n awoooi-prod -o jsonpath='{.data.OPENCLAW_URL}')",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8088/health)",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8088/)",
-      "Bash(nc -zv 192.168.0.188 8088 -w 5)",
-      "Bash(ping -c 2 192.168.0.188)",
-      "Bash(ping -c 2 192.168.70.202)",
-      "Bash(grep -n \"mapToDualState\" /Users/ogt/awoooi/apps/web/src/app/[locale]/page.tsx -A 30)",
-      "Bash(head -40 /Users/ogt/awoooi/apps/web/src/app/[locale]/page.tsx)",
-      "Bash(ssh -o ConnectTimeout=10 -o StrictHostKeyChecking=no ollama@192.168.0.188 \"docker ps -a | grep -i claw; docker start openclaw 2>/dev/null || docker start clawbot 2>/dev/null || echo ''Container not found, listing all:'' && docker ps -a --format ''table {{.Names}}\\\\t{{.Status}}'' | head -10\")",
-      "Bash(curl -s --connect-timeout 5 http://192.168.0.188:8089/health)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl rollout status deployment/awoooi-web -n awoooi-prod --timeout=60s)",
-      "Bash(grep -rn \"clawbot\\\\|ClawBot\" /Users/ogt/awoooi/ --include=*.yaml --include=*.yml --include=*.json)",
-      "Bash(grep -rn \"ClawBot\\\\|clawbot\" /Users/ogt/awoooi/apps/ --include=*.py --include=*.ts --include=*.tsx)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs deployment/awoooi-api -n awoooi-prod --tail=100)",
-      "Bash(KUBECONFIG=/Users/ogt/awoooi/apps/api/k3s-prod.yaml kubectl logs deployment/awoooi-api -n awoooi-prod --tail=200)",
-      "Bash(export KUBECONFIG=/Users/ogt/awoooi/k3s-prod.yaml)",
-      "Bash(ssh root@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=200 2>&1 | grep -iE ''error|fail|exception|execute|background|parse'' | tail -40\")",
-      "Bash(curl -s https://awoooi.wooo.work/api/v1/approvals)",
-      "Bash(ssh k3s@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=200 2>&1 | grep -iE ''error|fail|execute|background|parse'' | tail -40\")",
-      "Bash(ssh ubuntu@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=200 2>&1 | grep -iE ''error|fail|execute|background|parse'' | tail -40\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=200 2>&1 | grep -iE ''error|fail|execute|background|parse|skip'' | tail -50\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=500 2>&1 | grep -iE ''background_execution|approve_action|reject|k8s_executor'' | tail -30\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl get deploy,sts -n awoooi-prod\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl rollout status deployment/awoooi-api -n awoooi-prod --timeout=120s 2>&1\")",
-      "Bash(ssh wooo@192.168.0.120 \"kubectl logs deployment/awoooi-api -n awoooi-prod --tail=50 2>&1 | grep -iE ''background_execution|k8s_executor|parse'' | tail -10\")"
-    ],
-    "additionalDirectories": [
-      "/Users/ogt/awoooi/docs",
-      "/Users/ogt/.claude/projects/-Users-ogt-awoooi/memory",
-      "/Users/ogt/awoooi/apps/web/src/app",
-      "/Users/ogt/awoooi/apps/api",
-      "/Users/ogt/awoooi/apps/api/http:/localhost:8000/api/v1",
-      "/Users/ogt/awoooi/apps/web/public",
-      "/Users/ogt/Downloads",
-      "/Users/ogt/awoooi/apps/web/test-results",
-      "/Users/ogt/awoooi",
-      "/Users/ogt/awoooi/apps/web/src/app/[locale]",
-      "/tmp"
-    ]
-  }
-}

.dockerignore

@@ -19,10 +19,18 @@
 
 # 文件與腳本（不需要進 image）
 # 注意: docs/runbooks/, docs/adr/, .agents/skills/ 供 RAG 索引 (ADR-067 Phase 33)
-# scripts/ 大部分不需要進 image，但 CronJob 腳本需要
+# scripts/ 大部分不需要進 image，僅白名單 production runtime/ops 種子腳本
 # 2026-04-12 ogt (ADR-073 P2-1): 白名單允許 cron_km_vectorize.py
-scripts
+# 2026-05-13 codex: 白名單 T16 auto-repair canary PlayBook seed script
+# 2026-05-31 codex: MOMO backup Ansible playbook copies the backup script from
+# the controller image; keep only this backup script in the runtime context.
+scripts/**
+!scripts/
 !scripts/cron_km_vectorize.py
+!scripts/backup/
+!scripts/backup/backup-momo-188-pg.sh
+!scripts/ops/
+!scripts/ops/awooop-seed-auto-repair-canary-playbook.py
 
 # Node 快取（monorepo 根目錄）
 node_modules

.gitea/workflows/ansible-lint.yml

@@ -10,7 +10,7 @@ on:
 
 jobs:
   lint:
-    runs-on: self-hosted
+    runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
 

.gitea/workflows/cd-dev.yaml

@@ -43,10 +43,19 @@ jobs:
           ├ 📝 ${{ steps.commit.outputs.message }}
           ├ 🔖 <code>${{ steps.commit.outputs.short_sha }}</code>
           └ 🌿 dev branch"
-          printf '%b' "$MSG" | curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
-            -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
-            -d "parse_mode=HTML" \
-            --data-urlencode "text@-"
+          if AWOOI_CICD_STATUS=running \
+            AWOOI_CICD_STAGE=dev-deploy \
+            AWOOI_CICD_JOB_NAME="[DEV] 部署開始" \
+            AWOOI_CICD_COMMIT_SHA="${GITHUB_SHA}" \
+            AWOOI_CICD_SUMMARY="${{ steps.commit.outputs.message }}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "Dev deploy start notification mirrored through AWOOI API"
+          else
+            printf '%b' "$MSG" | curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
+              -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
+              -d "parse_mode=HTML" \
+              --data-urlencode "text@-"
+          fi
 
       # API 測試 (同 prod CI，確保 dev 也通過)
       - name: Run API Tests
@@ -78,11 +87,18 @@ jobs:
           echo "✅ API 測試通過"
 
       - name: Login to Harbor
-        uses: docker/login-action@v3
-        with:
-          registry: ${{ env.HARBOR }}
-          username: ${{ secrets.HARBOR_USERNAME }}
-          password: ${{ secrets.HARBOR_PASSWORD }}
+        run: |
+          HARBOR_USERNAME="$(cat <<'AWOOOI_SECRET_HARBOR_USERNAME'
+          ${{ secrets.HARBOR_USERNAME }}
+          AWOOOI_SECRET_HARBOR_USERNAME
+          )"
+          HARBOR_PASSWORD="$(cat <<'AWOOOI_SECRET_HARBOR_PASSWORD'
+          ${{ secrets.HARBOR_PASSWORD }}
+          AWOOOI_SECRET_HARBOR_PASSWORD
+          )"
+          printf '%s' "$HARBOR_PASSWORD" | docker login "${{ env.HARBOR }}" \
+            -u "$HARBOR_USERNAME" \
+            --password-stdin
 
       # Dev API 鏡像：強制重建，不用 cache（確保 models.json 等配置文件更新）
       - name: Build and Push API (Dev)
@@ -98,16 +114,37 @@ jobs:
 
       # 注入 Dev K8s Secrets
       - name: Inject Dev K8s Secrets
-        env:
-          SSH_PRIVATE_KEY: ${{ secrets.DEPLOY_SSH_KEY }}
-          TG_BOT_TOKEN: ${{ secrets.TELEGRAM_BOT_TOKEN }}
-          TG_CHAT_ID: ${{ secrets.TELEGRAM_CHAT_ID }}
-          NVIDIA_API_KEY: ${{ secrets.NVIDIA_API_KEY }}
-          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
         run: |
+          secret_b64() {
+            python3 -c 'import base64, sys; data=sys.stdin.buffer.read(); data=data[:-1] if data.endswith(b"\n") else data; sys.stdout.write(base64.b64encode(data).decode())'
+          }
+          write_deploy_key() {
+            mkdir -p ~/.ssh
+            umask 077
+            cat > ~/.ssh/deploy_key <<'AWOOOI_DEPLOY_KEY'
+          ${{ secrets.DEPLOY_SSH_KEY }}
+          AWOOOI_DEPLOY_KEY
+            chmod 600 ~/.ssh/deploy_key
+          }
+          TG_BOT_TOKEN_B64="$(secret_b64 <<'AWOOOI_SECRET_TG_BOT_TOKEN'
+          ${{ secrets.TELEGRAM_BOT_TOKEN }}
+          AWOOOI_SECRET_TG_BOT_TOKEN
+          )"
+          TG_CHAT_ID_B64="$(secret_b64 <<'AWOOOI_SECRET_TG_CHAT_ID'
+          ${{ secrets.TELEGRAM_CHAT_ID }}
+          AWOOOI_SECRET_TG_CHAT_ID
+          )"
+          NVIDIA_API_KEY_B64="$(secret_b64 <<'AWOOOI_SECRET_NVIDIA_API_KEY'
+          ${{ secrets.NVIDIA_API_KEY }}
+          AWOOOI_SECRET_NVIDIA_API_KEY
+          )"
+          GEMINI_API_KEY_B64="$(secret_b64 <<'AWOOOI_SECRET_GEMINI_API_KEY'
+          ${{ secrets.GEMINI_API_KEY }}
+          AWOOOI_SECRET_GEMINI_API_KEY
+          )"
+
           mkdir -p ~/.ssh
-          echo "$SSH_PRIVATE_KEY" > ~/.ssh/deploy_key
-          chmod 600 ~/.ssh/deploy_key
+          write_deploy_key
           # 2026-05-05 Codex: kubectl runs on 120 control-plane. 121 is a
           # worker and its local kubeconfig points at 127.0.0.1:6443.
           ssh -o StrictHostKeyChecking=no -i ~/.ssh/deploy_key wooo@192.168.0.120 << SECRETS
@@ -115,19 +152,19 @@ jobs:
           export KUBECONFIG=/etc/rancher/k3s/k3s.yaml
 
           sudo kubectl patch secret awoooi-secrets -n awoooi-dev --type='json' -p='[
-            {"op":"replace","path":"/data/OPENCLAW_TG_BOT_TOKEN","value":"'"$(echo -n "${TG_BOT_TOKEN}" | base64 -w 0)"'"},
-            {"op":"replace","path":"/data/OPENCLAW_TG_CHAT_ID","value":"'"$(echo -n "${TG_CHAT_ID}" | base64 -w 0)"'"}
+            {"op":"replace","path":"/data/OPENCLAW_TG_BOT_TOKEN","value":"${TG_BOT_TOKEN_B64}"},
+            {"op":"replace","path":"/data/OPENCLAW_TG_CHAT_ID","value":"${TG_CHAT_ID_B64}"}
           ]' || echo "⚠️ Telegram Secrets patch 跳過"
 
-          if [ -n "${NVIDIA_API_KEY}" ]; then
+          if [ -n "${NVIDIA_API_KEY_B64}" ]; then
             sudo kubectl patch secret awoooi-secrets -n awoooi-dev --type='json' -p='[
-              {"op":"replace","path":"/data/NVIDIA_API_KEY","value":"'"$(echo -n "${NVIDIA_API_KEY}" | base64 -w 0)"'"}
+              {"op":"replace","path":"/data/NVIDIA_API_KEY","value":"${NVIDIA_API_KEY_B64}"}
             ]' && echo "✅ NVIDIA_API_KEY 已注入 dev"
           fi
 
-          if [ -n "${GEMINI_API_KEY}" ]; then
+          if [ -n "${GEMINI_API_KEY_B64}" ]; then
             sudo kubectl patch secret awoooi-secrets -n awoooi-dev --type='json' -p='[
-              {"op":"replace","path":"/data/GEMINI_API_KEY","value":"'"$(echo -n "${GEMINI_API_KEY}" | base64 -w 0)"'"}
+              {"op":"replace","path":"/data/GEMINI_API_KEY","value":"${GEMINI_API_KEY_B64}"}
             ]' && echo "✅ GEMINI_API_KEY 已注入 dev"
           fi
 
@@ -136,8 +173,6 @@ jobs:
 
       # 部署到 awoooi-dev
       - name: Deploy to Dev K8s
-        env:
-          SSH_PRIVATE_KEY: ${{ secrets.DEPLOY_SSH_KEY }}
         run: |
           cat k8s/awoooi-dev/02-configmap.yaml | \
             ssh -o StrictHostKeyChecking=no -i ~/.ssh/deploy_key wooo@192.168.0.120 \
@@ -184,10 +219,20 @@ jobs:
           ├ 🔖 <code>${{ steps.commit.outputs.short_sha }}</code>
           ├ ⏱️ 耗時: ${MINUTES}m ${SECONDS}s
           └ 🩺 http://192.168.0.125:32344/api/v1/health"
-          printf '%b' "$MSG" | curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
-            -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
-            -d "parse_mode=HTML" \
-            --data-urlencode "text@-"
+          if AWOOI_CICD_STATUS=success \
+            AWOOI_CICD_STAGE=dev-deploy \
+            AWOOI_CICD_JOB_NAME="[DEV] 部署完成" \
+            AWOOI_CICD_COMMIT_SHA="${GITHUB_SHA}" \
+            AWOOI_CICD_DURATION_SECONDS="${DURATION}" \
+            AWOOI_CICD_SUMMARY="${{ steps.commit.outputs.message }}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "Dev deploy success notification mirrored through AWOOI API"
+          else
+            printf '%b' "$MSG" | curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
+              -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
+              -d "parse_mode=HTML" \
+              --data-urlencode "text@-"
+          fi
 
       - name: Notify Dev Deploy Failure
         if: failure()
@@ -196,7 +241,16 @@ jobs:
           ├ 📝 ${{ steps.commit.outputs.message }}
           ├ 🔖 <code>${{ steps.commit.outputs.short_sha }}</code>
           └ 🔗 <a href=\"http://192.168.0.110:3001/wooo/awoooi/actions\">查看日誌</a>"
-          printf '%b' "$MSG" | curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
-            -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
-            -d "parse_mode=HTML" \
-            --data-urlencode "text@-"
+          if AWOOI_CICD_STATUS=failed \
+            AWOOI_CICD_STAGE=dev-deploy \
+            AWOOI_CICD_JOB_NAME="[DEV] 部署失敗" \
+            AWOOI_CICD_COMMIT_SHA="${GITHUB_SHA}" \
+            AWOOI_CICD_SUMMARY="${{ steps.commit.outputs.message }}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "Dev deploy failure notification mirrored through AWOOI API"
+          else
+            printf '%b' "$MSG" | curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
+              -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
+              -d "parse_mode=HTML" \
+              --data-urlencode "text@-"
+          fi

.gitea/workflows/code-review.yaml

@@ -30,6 +30,9 @@ jobs:
         with:
           fetch-depth: 50
 
+      - name: Guard Workflow Secret Surfaces
+        run: node scripts/ci/check-gitea-step-env-secrets.js
+
       - name: Skip Stale Main Push
         id: stale
         run: |
@@ -102,7 +105,6 @@ jobs:
       - name: Notify Code Review Start
         if: steps.stale.outputs.skip != 'true'
         env:
-          TG_BOT_TOKEN: ${{ secrets.TELEGRAM_BOT_TOKEN }}
           TG_CHAT_ID: ${{ env.TELEGRAM_ALERT_CHAT_ID }}
           SHORT_SHA: ${{ steps.ctx.outputs.short_sha }}
           BRANCH: ${{ steps.ctx.outputs.branch }}
@@ -110,18 +112,33 @@ jobs:
           FILES_DISPLAY: ${{ steps.ctx.outputs.files_display }}
         run: |
           set -euo pipefail
-          if [ -z "${TG_BOT_TOKEN:-}" ] || [ -z "${TG_CHAT_ID:-}" ]; then
-            echo "Telegram secret missing; skip start notification"
-            exit 0
-          fi
+          TG_BOT_TOKEN="$(cat <<'AWOOOI_SECRET_TG_BOT_TOKEN'
+          ${{ secrets.TELEGRAM_BOT_TOKEN }}
+          AWOOOI_SECRET_TG_BOT_TOKEN
+          )"
           html_escape() { sed 's/&/\&amp;/g; s/</\&lt;/g; s/>/\&gt;/g'; }
           COMMIT_ESC="$(printf '%s' "$COMMIT_MSG" | html_escape)"
           FILES_ESC="$(printf '%s\n' "$FILES_DISPLAY" | html_escape)"
           MSG="$(printf '🔍 <b>Code Review 啟動</b>\n──────────────────────\n📦 Commit <code>%s</code> 🌿 <code>%s</code>\n📝 <code>%s</code>\n📁 <b>變更檔案：</b>\n%s\n──────────────────────\n🤖 <b>Hermes → OpenClaw → Elephant Alpha → NemoTron</b>\n📊 即時進度：<a href=\"%s\">%s</a>' "$SHORT_SHA" "$BRANCH" "$COMMIT_ESC" "$FILES_ESC" "$REPORT_URL" "$REPORT_URL")"
-          curl -fsS -X POST "https://api.telegram.org/bot${TG_BOT_TOKEN}/sendMessage" \
-            -H "Content-Type: application/json" \
-            -d "$(jq -n --arg c "$TG_CHAT_ID" --arg t "$MSG" '{chat_id:$c,text:$t,parse_mode:"HTML",disable_web_page_preview:true}')" \
-            >/dev/null
+          if AWOOI_CICD_STATUS=running \
+            AWOOI_CICD_STAGE=code-review \
+            AWOOI_CICD_JOB_NAME="Code Review 啟動" \
+            AWOOI_CICD_COMMIT_SHA="${GITHUB_SHA}" \
+            AWOOI_CICD_TRIGGERED_BY="${GITHUB_ACTOR:-CI}" \
+            AWOOI_CICD_SUMMARY="${COMMIT_MSG}" \
+            AWOOI_CICD_WORKFLOW_URL="${REPORT_URL}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "Code review start notification mirrored through AWOOI API"
+          else
+            if [ -z "${TG_BOT_TOKEN:-}" ] || [ -z "${TG_CHAT_ID:-}" ]; then
+              echo "Telegram secret missing and AWOOI API notify failed; skip start notification"
+              exit 0
+            fi
+            curl -fsS -X POST "https://api.telegram.org/bot${TG_BOT_TOKEN}/sendMessage" \
+              -H "Content-Type: application/json" \
+              -d "$(jq -n --arg c "$TG_CHAT_ID" --arg t "$MSG" '{chat_id:$c,text:$t,parse_mode:"HTML",disable_web_page_preview:true}')" \
+              >/dev/null
+          fi
 
       - name: Run Deterministic Review
         if: steps.stale.outputs.skip != 'true'
@@ -139,15 +156,14 @@ jobs:
       - name: Notify Code Review Completion
         if: always() && steps.stale.outputs.skip != 'true'
         env:
-          TG_BOT_TOKEN: ${{ secrets.TELEGRAM_BOT_TOKEN }}
           TG_CHAT_ID: ${{ env.TELEGRAM_ALERT_CHAT_ID }}
           SHORT_SHA: ${{ steps.ctx.outputs.short_sha }}
         run: |
           set -euo pipefail
-          if [ -z "${TG_BOT_TOKEN:-}" ] || [ -z "${TG_CHAT_ID:-}" ]; then
-            echo "Telegram secret missing; skip completion notification"
-            exit 0
-          fi
+          TG_BOT_TOKEN="$(cat <<'AWOOOI_SECRET_TG_BOT_TOKEN'
+          ${{ secrets.TELEGRAM_BOT_TOKEN }}
+          AWOOOI_SECRET_TG_BOT_TOKEN
+          )"
           REPORT=/tmp/code-review-report.json
           if [ ! -s "$REPORT" ]; then
             cat > "$REPORT" <<'JSON'
@@ -180,7 +196,25 @@ jobs:
           TOP_ESC="$(printf '%s' "$TOP_ISSUE" | html_escape)"
 
           MSG="$(printf '%s <b>Code Review 完成・%s</b>\n──────────────────────\n🔴 CRITICAL <code>%s</code>  🟠 HIGH <code>%s</code>  🟡 MEDIUM <code>%s</code>  🟢 LOW <code>%s</code>\n──────────────────────\n⚠️ <b>主要問題</b>\n%s\n\n🔍 <b>整體風險等級</b>\n%s：%s\n\n⚠️ <b>最高關注問題</b>\n1. %s\n──────────────────────\n🤖 Elephant Alpha：<b>%s</b> ✅ %s\n📊 完整報告：<a href=\"%s\">%s</a>' "$STATUS" "$SHORT_SHA" "$CRITICAL" "$HIGH" "$MEDIUM" "$LOW" "$ISSUE_LINE" "$RISK" "$SUMMARY_ESC" "$TOP_ESC" "$RISK" "$ACTION_ESC" "$REPORT_URL" "$REPORT_URL")"
-          curl -fsS -X POST "https://api.telegram.org/bot${TG_BOT_TOKEN}/sendMessage" \
-            -H "Content-Type: application/json" \
-            -d "$(jq -n --arg c "$TG_CHAT_ID" --arg t "$MSG" '{chat_id:$c,text:$t,parse_mode:"HTML",disable_web_page_preview:true}')" \
-            >/dev/null
+          CICD_STATUS=success
+          if [ "$RISK" = "MEDIUM" ]; then CICD_STATUS=pending; fi
+          if [ "$RISK" = "HIGH" ] || [ "$RISK" = "CRITICAL" ]; then CICD_STATUS=failed; fi
+          if AWOOI_CICD_STATUS="${CICD_STATUS}" \
+            AWOOI_CICD_STAGE=code-review \
+            AWOOI_CICD_JOB_NAME="Code Review 完成・${RISK}" \
+            AWOOI_CICD_COMMIT_SHA="${GITHUB_SHA}" \
+            AWOOI_CICD_TRIGGERED_BY="${GITHUB_ACTOR:-CI}" \
+            AWOOI_CICD_SUMMARY="CRITICAL=${CRITICAL}; HIGH=${HIGH}; MEDIUM=${MEDIUM}; LOW=${LOW}; ${SUMMARY}" \
+            AWOOI_CICD_WORKFLOW_URL="${REPORT_URL}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "Code review completion notification mirrored through AWOOI API"
+          else
+            if [ -z "${TG_BOT_TOKEN:-}" ] || [ -z "${TG_CHAT_ID:-}" ]; then
+              echo "Telegram secret missing and AWOOI API notify failed; skip completion notification"
+              exit 0
+            fi
+            curl -fsS -X POST "https://api.telegram.org/bot${TG_BOT_TOKEN}/sendMessage" \
+              -H "Content-Type: application/json" \
+              -d "$(jq -n --arg c "$TG_CHAT_ID" --arg t "$MSG" '{chat_id:$c,text:$t,parse_mode:"HTML",disable_web_page_preview:true}')" \
+              >/dev/null
+          fi

.gitea/workflows/deploy-alerts.yaml

@@ -1,7 +1,7 @@
 # =============================================================================
 # Deploy Prometheus Alert Rules (獨立 workflow)
 # 2026-04-05 Claude Code (ADR-039 I3): 從 cd.yaml 分離
-# 觸發條件: ops/monitoring/alerts-unified.yml 有變更 或 workflow_dispatch
+# 觸發條件: ops/monitoring/alerts-unified.yml / slo-rules.yml 有變更 或 workflow_dispatch
 # 說明: 告警規則部署不依賴應用構建，獨立觸發以加快響應速度
 # =============================================================================
 
@@ -12,6 +12,8 @@ on:
     branches: [main]
     paths:
       - 'ops/monitoring/alerts-unified.yml'
+      - 'ops/monitoring/slo-rules.yml'
+      - 'scripts/ops/deploy-alerts.sh'
   workflow_dispatch:
 
 env:
@@ -30,11 +32,15 @@ jobs:
         run: |
           pip3 install -q pyyaml 2>/dev/null || pip install -q pyyaml
           python3 -c "import yaml; yaml.safe_load(open('ops/monitoring/alerts-unified.yml')); print('YAML OK')"
+          python3 -c "import yaml; yaml.safe_load(open('ops/monitoring/slo-rules.yml')); print('SLO YAML OK')"
 
       - name: Setup SSH key
         run: |
           mkdir -p ~/.ssh
-          echo "${{ secrets.DEPLOY_SSH_KEY }}" > ~/.ssh/id_ed25519
+          umask 077
+          cat > ~/.ssh/id_ed25519 <<'AWOOOI_DEPLOY_KEY'
+          ${{ secrets.DEPLOY_SSH_KEY }}
+          AWOOOI_DEPLOY_KEY
           chmod 600 ~/.ssh/id_ed25519
           ssh-keyscan 192.168.0.110 >> ~/.ssh/known_hosts
 
@@ -50,6 +56,17 @@ jobs:
           SHORT_SHA="${{ github.sha }}"
           SHORT_SHA="${SHORT_SHA:0:7}"
           MSG="${EMOJI} Prometheus 告警規則部署 ${STATUS} (${SHORT_SHA})"
-          curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
-            -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
-            --data-urlencode "text=${MSG}" || true
+          CICD_STATUS="success"
+          [ "$STATUS" != "success" ] && CICD_STATUS="failed"
+          if AWOOI_CICD_STATUS="${CICD_STATUS}" \
+            AWOOI_CICD_STAGE=deploy-alerts \
+            AWOOI_CICD_JOB_NAME="Prometheus 告警規則部署" \
+            AWOOI_CICD_COMMIT_SHA="${{ github.sha }}" \
+            AWOOI_CICD_SUMMARY="${MSG}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "Alert rule deploy notification mirrored through AWOOI API"
+          else
+            curl -fS -X POST "https://api.telegram.org/bot${{ secrets.TELEGRAM_BOT_TOKEN }}/sendMessage" \
+              -d "chat_id=${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
+              --data-urlencode "text=${MSG}" || true
+          fi

.gitea/workflows/e2e-health.yaml

@@ -51,10 +51,52 @@ jobs:
           echo "status=failed" >> $GITHUB_OUTPUT
           exit 1
 
+      - name: Source Provider Freshness Smoke
+        run: |
+          SOURCE_CANARY_RUN_REF="gitea-e2e-${GITHUB_RUN_ID:-manual}-${GITHUB_RUN_ATTEMPT:-1}"
+          echo "SOURCE_CANARY_RUN_REF=${SOURCE_CANARY_RUN_REF}" >> "$GITHUB_ENV"
+          echo "SOURCE_LINK_CANARY_WORK_ITEM_ID=source-evidence:sentry:upstream_canary:awoooi-source-link-canary-${SOURCE_CANARY_RUN_REF}" >> "$GITHUB_ENV"
+          OPERATOR_KEY="$(cat <<'AWOOOI_SECRET_AWOOOP_OPERATOR_API_KEY'
+          ${{ secrets.AWOOOP_OPERATOR_API_KEY }}
+          AWOOOI_SECRET_AWOOOP_OPERATOR_API_KEY
+          )"
+          AWOOOP_OPERATOR_API_KEY="${OPERATOR_KEY}" \
+          AWOOOP_OPERATOR_ID=gitea-e2e-health \
+          python3 scripts/alert_chain_smoke_test.py \
+            --api-url https://awoooi.wooo.work \
+            --metrics-api-url http://192.168.0.125:32334 \
+            --source-provider-heartbeat \
+            --source-provider-upstream-canary \
+            --run-ref "${SOURCE_CANARY_RUN_REF}" \
+            --source-link-canary-target-incident-id INC-20260505-25E744 \
+            --json
+
+      - name: Source Correlation Applied-Link Smoke
+        run: |
+          python3 scripts/awooop_source_correlation_apply_smoke.py \
+            --api-url https://awoooi.wooo.work \
+            --target-incident-id INC-20260505-25E744 \
+            --allow-existing-apply \
+            --refresh-if-stale-days 6 \
+            --refresh-work-item-id "${SOURCE_LINK_CANARY_WORK_ITEM_ID}" \
+            --verify-refresh-candidate \
+            --reviewer-id gitea_e2e_source_link_canary \
+            --operator-note "T124 dedicated source-link canary refresh; append-only status-chain proof"
+
       - name: Notify Telegram on Failure
         if: failure()
         run: |
-          curl -s -X POST "https://api.telegram.org/bot${{ secrets.OPENCLAW_TG_BOT_TOKEN }}/sendMessage" \
-            -d chat_id="${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
-            -d parse_mode="HTML" \
-            -d text="🔴 <b>[E2E Health Check]</b> 失敗%0A%0A📅 $(TZ=Asia/Taipei date '+%Y-%m-%d %H:%M')%0A🔗 API 健康檢查未通過%0A%0A請檢查 K3s 叢集狀態"
+          MSG="E2E Health Check 失敗；API 健康檢查未通過"
+          if AWOOI_CICD_STATUS=failed \
+            AWOOI_CICD_STAGE=e2e-health \
+            AWOOI_CICD_JOB_NAME="E2E Health Check" \
+            AWOOI_CICD_COMMIT_SHA="${{ github.sha }}" \
+            AWOOI_CICD_SUMMARY="${MSG}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "E2E failure notification mirrored through AWOOI API"
+          else
+            curl -s -X POST "https://api.telegram.org/bot${{ secrets.OPENCLAW_TG_BOT_TOKEN }}/sendMessage" \
+              -d chat_id="${{ env.TELEGRAM_ALERT_CHAT_ID }}" \
+              -d parse_mode="HTML" \
+              -d text="🔴 <b>[E2E Health Check]</b> 失敗%0A%0A📅 $(TZ=Asia/Taipei date '+%Y-%m-%d %H:%M')%0A🔗 API 健康檢查未通過%0A%0A請檢查 K3s 叢集狀態"
+          fi

.gitea/workflows/run-migration.yml

@@ -57,23 +57,34 @@ jobs:
       - name: Identify new migrations
         id: diff
         run: |
-          NEW_FILES=$(git diff --name-only --diff-filter=A HEAD~1 HEAD -- 'apps/api/migrations/*.sql' || true)
+          ALL_NEW_FILES=$(git diff --no-renames --name-only --diff-filter=A HEAD~1 HEAD -- 'apps/api/migrations/*.sql' || true)
+          NEW_FILES=$(echo "$ALL_NEW_FILES" | grep -Ev '(_down|rollback)\.sql$' || true)
+          SKIPPED_ROLLBACK_FILES=$(echo "$ALL_NEW_FILES" | grep -E '(_down|rollback)\.sql$' || true)
           echo "new_files<<EOF" >> $GITHUB_OUTPUT
           echo "$NEW_FILES" >> $GITHUB_OUTPUT
           echo "EOF" >> $GITHUB_OUTPUT
           echo "=== New migration files ==="
           echo "$NEW_FILES"
+          if [ -n "$SKIPPED_ROLLBACK_FILES" ]; then
+            echo "=== Rollback/down migrations skipped by design ==="
+            echo "$SKIPPED_ROLLBACK_FILES"
+          fi
 
       - name: Apply new migrations
         if: steps.diff.outputs.new_files != ''
-        env:
-          # 從 Gitea secrets 取，不直接明碼輸出。
-          # MIGRATION_DATABASE_URL 是限權帳號；DATABASE_URL 只在 PostgreSQL
-          # 明確回報「必須是 table owner」時作為受控 fallback。
-          PGURL: ${{ secrets.MIGRATION_DATABASE_URL }}
-          OWNER_PGURL: ${{ secrets.DATABASE_URL }}
         run: |
           set -euo pipefail
+          # 從 Gitea secrets 取，不放 step-level env，避免 runner log 展開。
+          # MIGRATION_DATABASE_URL 是限權帳號；DATABASE_URL 只在 PostgreSQL
+          # 明確回報「必須是 table owner」時作為受控 fallback。
+          PGURL="$(cat <<'AWOOOI_SECRET_MIGRATION_DATABASE_URL'
+          ${{ secrets.MIGRATION_DATABASE_URL }}
+          AWOOOI_SECRET_MIGRATION_DATABASE_URL
+          )"
+          OWNER_PGURL="$(cat <<'AWOOOI_SECRET_DATABASE_URL'
+          ${{ secrets.DATABASE_URL }}
+          AWOOOI_SECRET_DATABASE_URL
+          )"
           if [ -z "$PGURL" ]; then
             echo "::error::MIGRATION_DATABASE_URL secret not set in Gitea"
             exit 1
@@ -96,7 +107,7 @@ jobs:
             echo "=== Applying: $file ==="
             migration_err="$(mktemp)"
             if ! apply_migration "$PGURL_PSQL" "$file" 2>"$migration_err"; then
-              if grep -q "must be owner of table" "$migration_err"; then
+              if grep -Eq "(must be owner of table|permission denied for table)" "$migration_err"; then
                 if [ -z "$OWNER_PGURL_PSQL" ]; then
                   cat "$migration_err" >&2
                   echo "::error::migration requires table owner but DATABASE_URL secret is not set"
@@ -115,11 +126,16 @@ jobs:
 
       - name: Seed asset_discovery_run (audit)
         if: steps.diff.outputs.new_files != ''
-        env:
-          PGURL: ${{ secrets.MIGRATION_DATABASE_URL }}
-          OWNER_PGURL: ${{ secrets.DATABASE_URL }}
         run: |
           set -euo pipefail
+          PGURL="$(cat <<'AWOOOI_SECRET_MIGRATION_DATABASE_URL'
+          ${{ secrets.MIGRATION_DATABASE_URL }}
+          AWOOOI_SECRET_MIGRATION_DATABASE_URL
+          )"
+          OWNER_PGURL="$(cat <<'AWOOOI_SECRET_DATABASE_URL'
+          ${{ secrets.DATABASE_URL }}
+          AWOOOI_SECRET_DATABASE_URL
+          )"
           if [ -z "$PGURL" ]; then
             echo "::error::MIGRATION_DATABASE_URL secret not set in Gitea"
             exit 1
@@ -127,14 +143,15 @@ jobs:
           PGURL_PSQL="${PGURL/postgresql+asyncpg:\/\//postgresql:\/\/}"
           OWNER_PGURL_PSQL="${OWNER_PGURL/postgresql+asyncpg:\/\//postgresql:\/\/}"
           FILES_JSON=$(echo "${{ steps.diff.outputs.new_files }}" | jq -Rn '[inputs | select(length > 0)]')
+          SUMMARY_JSON=$(jq -cn \
+            --arg commit_sha "${{ github.sha }}" \
+            --argjson files "$FILES_JSON" \
+            '{type: "ci_migration", commit_sha: $commit_sha, files: $files}')
+          SUMMARY_JSON_SQL=${SUMMARY_JSON//\'/\'\'}
 
           seed_audit() {
             local url="$1"
-            psql "$url" \
-              -v ON_ERROR_STOP=1 \
-              -v commit_sha="${{ github.sha }}" \
-              -v files_json="$FILES_JSON" \
-              -c "
+            psql "$url" -v ON_ERROR_STOP=1 <<SQL
             INSERT INTO asset_discovery_run (
               run_id, triggered_by, scope, scan_depth, status,
               started_at, ended_at, tools_used, summary
@@ -146,14 +163,10 @@ jobs:
               'success',
               NOW(),
               NOW(),
-              '{\"psql\": 1, \"gitea_ci\": 1}'::jsonb,
-              jsonb_build_object(
-                'type', 'ci_migration',
-                'commit_sha', :'commit_sha',
-                'files', :'files_json'::jsonb
-              )
+              '{"psql": 1, "gitea_ci": 1}'::jsonb,
+              '${SUMMARY_JSON_SQL}'::jsonb
             );
-          "
+          SQL
           }
 
           audit_err="$(mktemp)"
@@ -176,11 +189,25 @@ jobs:
       - name: Notify Telegram (if configured)
         if: always()
         env:
-          TG_TOKEN: ${{ secrets.TELEGRAM_BOT_TOKEN }}
           TG_CHAT: ${{ env.TELEGRAM_ALERT_CHAT_ID }}
         run: |
+          TG_TOKEN="$(cat <<'AWOOOI_SECRET_TG_TOKEN'
+          ${{ secrets.TELEGRAM_BOT_TOKEN }}
+          AWOOOI_SECRET_TG_TOKEN
+          )"
+          STATUS="${{ job.status }}"
+          CICD_STATUS="success"
+          [ "$STATUS" != "success" ] && CICD_STATUS="failed"
+          if AWOOI_CICD_STATUS="${CICD_STATUS}" \
+            AWOOI_CICD_STAGE=run-migration \
+            AWOOI_CICD_JOB_NAME="Migration CI" \
+            AWOOI_CICD_COMMIT_SHA="${{ github.sha }}" \
+            AWOOI_CICD_SUMMARY="Migration CI: ${STATUS}" \
+            scripts/ci/notify-awoooi-cicd.sh; then
+            echo "Migration notification mirrored through AWOOI API"
+            exit 0
+          fi
           if [ -n "$TG_TOKEN" ] && [ -n "$TG_CHAT" ]; then
-            STATUS="${{ job.status }}"
             MSG="🗄️ Migration CI: \`${STATUS}\` — commit ${{ github.sha }}"
             curl -s -X POST "https://api.telegram.org/bot${TG_TOKEN}/sendMessage" \
               -d chat_id="${TG_CHAT}" \

.github/workflows/cd.yaml

@@ -13,12 +13,10 @@
 
 name: CD
 
+# 2026-05-12 Codex: GitHub 僅保留唯讀備份；生產 CI/CD 只能從 Gitea 執行。
+# 本 workflow 曾可 push / workflow_dispatch 後 build、patch secret、kubectl apply，
+# 會和 `.gitea/workflows/cd.yaml` 競爭 K3s production 狀態，因此硬停用。
 on:
-  push:
-    branches: [main]
-    paths-ignore:
-      - 'docs/**'
-      - '*.md'
   workflow_dispatch:
     inputs:
       force_deploy:
@@ -60,6 +58,7 @@ jobs:
   # ==================== Pre-flight Check (10s Fail-Fast) ====================
   pre-flight-check:
     name: "Pre-flight Check"
+    if: ${{ false }}
     runs-on: [self-hosted, harbor, k8s]
     timeout-minutes: 1
     steps:
@@ -133,6 +132,7 @@ jobs:
   # 2026-03-29 Claude Code: 確保監控覆蓋率 >= 90%
   monitoring-coverage:
     name: "Monitoring Coverage"
+    if: ${{ false }}
     runs-on: [self-hosted, harbor, k8s]
     needs: pre-flight-check
     timeout-minutes: 2
@@ -152,6 +152,7 @@ jobs:
   # ==================== 路徑偵測 (使用 dorny/paths-filter) ====================
   detect-changes:
     name: Detect Changes
+    if: ${{ false }}
     runs-on: [self-hosted, harbor, k8s]
     needs: [pre-flight-check, monitoring-coverage]
     timeout-minutes: 1
@@ -197,11 +198,7 @@ jobs:
     runs-on: [self-hosted, harbor, k8s]
     needs: [detect-changes, build-web]
     timeout-minutes: 20
-    if: |
-      !inputs.skip_api && (
-        needs.detect-changes.outputs.api == 'true' ||
-        (needs.detect-changes.outputs.api == 'false' && needs.detect-changes.outputs.web == 'false')
-      )
+    if: ${{ false }}
     outputs:
       image_tag: ${{ steps.tag.outputs.tag }}
     steps:
@@ -238,11 +235,7 @@ jobs:
     runs-on: [self-hosted, harbor, k8s]
     needs: detect-changes
     timeout-minutes: 20
-    if: |
-      !inputs.skip_web && (
-        needs.detect-changes.outputs.web == 'true' ||
-        (needs.detect-changes.outputs.api == 'false' && needs.detect-changes.outputs.web == 'false')
-      )
+    if: ${{ false }}
     outputs:
       image_tag: ${{ steps.tag.outputs.tag }}
     steps:
@@ -293,7 +286,7 @@ jobs:
     concurrency:
       group: runner-awoooi-cd-mutex
       cancel-in-progress: false
-    if: always() && (needs.build-api.result == 'success' || needs.build-api.result == 'skipped') && (needs.build-web.result == 'success' || needs.build-web.result == 'skipped')
+    if: ${{ false }}
     environment: production
     steps:
       # 2026-03-29: Runner 診斷檔案清理 (防止並行衝突)

.github/workflows/deploy-prod.yml

@@ -14,15 +14,10 @@
 
 name: Deploy to Production
 
+# 2026-05-12 Codex: GitHub 是唯讀備份，production deploy 只能從 Gitea 進入。
+# 這份歷史 workflow 仍含 Harbor build/push 與 kubectl apply/rollout，會和 Gitea CD 競爭。
+# 保留檔案供稽核，但停用所有 job。
 on:
-  push:
-    branches:
-      - main
-    paths:
-      - 'apps/api/**'
-      - 'apps/web/**'
-      - 'k8s/awoooi-prod/**'
-      - '.github/workflows/deploy-prod.yml'
   workflow_dispatch:
     inputs:
       deploy_api:
@@ -70,6 +65,7 @@ jobs:
   # ===========================================================================
   build:
     name: "Build Images"
+    if: ${{ false }}
     runs-on: [self-hosted, harbor, k8s]
     outputs:
       image_tag: ${{ steps.meta.outputs.tag }}
@@ -138,6 +134,7 @@ jobs:
   deploy:
     name: "Deploy to K3s"
     needs: build
+    if: ${{ false }}
     runs-on: [self-hosted, harbor, k8s]
 
     steps:
@@ -210,7 +207,7 @@ jobs:
   smoke-test:
     name: "Smoke Tests"
     needs: deploy
-    if: ${{ !inputs.skip_tests }}
+    if: ${{ false }}
     runs-on: [self-hosted, harbor, k8s]
 
     steps:
@@ -248,7 +245,7 @@ jobs:
   notify:
     name: "Send Notification"
     needs: [build, deploy, smoke-test]
-    if: always()
+    if: ${{ false }}
     runs-on: [self-hosted, harbor, k8s]
 
     steps:

.gitignore

@@ -93,3 +93,4 @@ tsconfig.tsbuildinfo
 !.aiderignore
 .claude/settings.local.json
 .claude/settings.json
+.claude/settings.json.bak*

apps/api/.cd-trigger

@@ -1 +1 @@
-# 2026-04-05 warm-up deploy triggered
+# 2026-05-20 source-provider-heartbeat deploy trigger

apps/api/Dockerfile

@@ -44,28 +44,6 @@ FROM python:3.11-slim
 
 WORKDIR /app
 
-# Copy installed packages from builder
-COPY --from=builder /usr/local/lib/python3.11/site-packages /usr/local/lib/python3.11/site-packages
-COPY --from=builder /usr/local/bin /usr/local/bin
-
-# 2026-04-01 ogt: CACHE_BUST 強制失效 src/ 和 models.json 層
-# deps 層 (pip install) 仍可 cache；代碼/配置變更必須重建
-ARG CACHE_BUST=none
-COPY apps/api/src/ ./src/
-COPY apps/api/models.json ./models.json
-# 2026-04-09 ogt: 規則引擎配置 — alert_rule_engine.py 從此檔載入規則
-COPY apps/api/alert_rules.yaml ./alert_rules.yaml
-# 2026-04-10 Claude Sonnet 4.6: drift_detector 需要 k8s/ YAML 做 Git state 比對
-COPY k8s/ ./k8s/
-# 2026-04-10 Claude Sonnet 4.6: RAG 知識庫索引來源 (ADR-067 Phase 33)
-COPY docs/ ./docs/
-COPY .agents/skills/ ./.agents/skills/
-# 2026-05-04 Claude Sonnet 4.6 (Task 1.2): hermes agent_loader 的 system prompt 來源
-# agent_loader.py 預設讀 /app/.claude/agents/，對應 K8s AGENTS_DIR 環境變數
-COPY .claude/agents/ ./.claude/agents/
-# 2026-04-12 ogt (ADR-073 P2-1): CronJob 腳本 — 獨立腳本取代 inline Python
-COPY scripts/ ./scripts/
-
 # Install openssh-client + curl — SSH_COMMAND Playbook + healthcheck
 # Install kubectl — drift_detector 需要 kubectl 讀取 K8s 實際狀態
 # (2026-04-09 Claude Sonnet 4.6 Asia/Taipei, Bug #6 修正 — python:3.11-slim 無 openssh-client)
@@ -75,8 +53,38 @@ RUN apt-get update && apt-get install -y --no-install-recommends openssh-client
     chmod +x kubectl && mv kubectl /usr/local/bin/kubectl && \
     rm -rf /var/lib/apt/lists/*
 
-# Create non-root user
-RUN useradd -m -u 1000 appuser && chown -R appuser:appuser /app
+# Create non-root user before copying app artifacts so COPY --chown can avoid
+# an expensive full-tree chown layer on every source-only rebuild.
+RUN useradd -m -u 1000 appuser
+
+# Copy installed packages from builder
+COPY --from=builder /usr/local/lib/python3.11/site-packages /usr/local/lib/python3.11/site-packages
+COPY --from=builder /usr/local/bin /usr/local/bin
+
+# 2026-04-01 ogt: CACHE_BUST 強制失效 src/ 和 models.json 層
+# deps 層 (pip install) 仍可 cache；代碼/配置變更必須重建
+ARG CACHE_BUST=none
+COPY --chown=appuser:appuser apps/api/src/ ./src/
+# 2026-04-09 ogt: 規則引擎配置 — alert_rule_engine.py 從此檔載入規則
+COPY --chown=appuser:appuser apps/api/models.json ./models.json
+COPY --chown=appuser:appuser apps/api/alert_rules.yaml ./alert_rules.yaml
+# 2026-04-10 Claude Sonnet 4.6: drift_detector 需要 k8s/ YAML 做 Git state 比對
+COPY --chown=appuser:appuser k8s/ ./k8s/
+# 2026-05-24 Codex: truth-chain / Ansible readiness needs the repo-known
+# playbook catalog in the API image.
+# 2026-05-31 Codex: ansible-core is now installed through pyproject.toml so
+# this catalog can graduate from visibility-only to check-mode runtime-ready
+# once repair SSH material is mounted and readable. This still does not enable
+# automatic apply; approval/execution code remains the gate.
+COPY --chown=appuser:appuser infra/ansible/ ./infra/ansible/
+# 2026-04-10 Claude Sonnet 4.6: RAG 知識庫索引來源 (ADR-067 Phase 33)
+COPY --chown=appuser:appuser docs/ ./docs/
+COPY --chown=appuser:appuser .agents/skills/ ./.agents/skills/
+# 2026-05-04 Claude Sonnet 4.6 (Task 1.2): hermes agent_loader 的 system prompt 來源
+# agent_loader.py 預設讀 /app/.claude/agents/，對應 K8s AGENTS_DIR 環境變數
+COPY --chown=appuser:appuser .claude/agents/ ./.claude/agents/
+# 2026-04-12 ogt (ADR-073 P2-1): CronJob 腳本 — 獨立腳本取代 inline Python
+COPY --chown=appuser:appuser scripts/ ./scripts/
 USER appuser
 
 # Expose port

apps/api/migrations/adr090d_ansible_operation_types.sql

@@ -0,0 +1,36 @@
+-- ADR-090-D: automation_operation_log.operation_type adds Ansible executor audit states
+-- Created: 2026-05-12 Taipei
+--
+-- Purpose:
+--   T3 Ansible declarative executor visibility. These operation types allow
+--   the AI automation truth chain to record that Ansible was matched,
+--   check-mode executed, applied, rolled back, or explicitly skipped.
+--
+-- Safety:
+--   This migration only expands the CHECK allowlist. It does not execute
+--   Ansible, change approval behavior, or create auto-remediation rows.
+
+ALTER TABLE automation_operation_log
+    DROP CONSTRAINT IF EXISTS automation_operation_log_type_valid;
+
+ALTER TABLE automation_operation_log
+    ADD CONSTRAINT automation_operation_log_type_valid CHECK (operation_type IN (
+        'monitor_configured','monitor_removed',
+        'alert_fired','alert_suppressed','alert_routed',
+        'rule_created','rule_updated','rule_matched','rule_rejected','rule_deprecated',
+        'playbook_generated','playbook_updated','playbook_executed',
+        'remediation_executed','remediation_verified','remediation_rolled_back',
+        'self_correction_attempted',
+        'km_created','km_updated','km_linked',
+        'asset_discovered','coverage_recalculated',
+        'capacity_recommendation','quota_enforced',
+        'notification_formatted',
+        'ansible_candidate_matched',
+        'ansible_check_mode_executed',
+        'ansible_apply_executed',
+        'ansible_rollback_executed',
+        'ansible_execution_skipped'
+    ));
+
+COMMENT ON CONSTRAINT automation_operation_log_type_valid ON automation_operation_log IS
+    'ADR-090-D: allow first-class Ansible executor audit states for AwoooP truth-chain visibility.';

apps/api/migrations/adr090d_ansible_operation_types_down.sql

@@ -0,0 +1,19 @@
+-- ADR-090-D rollback: remove Ansible executor audit states from operation_type allowlist.
+-- Only apply after confirming no automation_operation_log rows use ansible_* operation types.
+
+ALTER TABLE automation_operation_log
+    DROP CONSTRAINT IF EXISTS automation_operation_log_type_valid;
+
+ALTER TABLE automation_operation_log
+    ADD CONSTRAINT automation_operation_log_type_valid CHECK (operation_type IN (
+        'monitor_configured','monitor_removed',
+        'alert_fired','alert_suppressed','alert_routed',
+        'rule_created','rule_updated','rule_matched','rule_rejected','rule_deprecated',
+        'playbook_generated','playbook_updated','playbook_executed',
+        'remediation_executed','remediation_verified','remediation_rolled_back',
+        'self_correction_attempted',
+        'km_created','km_updated','km_linked',
+        'asset_discovered','coverage_recalculated',
+        'capacity_recommendation','quota_enforced',
+        'notification_formatted'
+    ));

apps/api/migrations/awooop_awoooi_mcp_approval_executor_ssh_gateway_2026-05-13.sql

@@ -0,0 +1,164 @@
+-- T9: approved SSH execution MCP Gateway seed
+-- 目的：讓 Telegram/Approval 已批准的 SSH 修復動作通過 AwoooP Gateway 五閘門。
+-- 邊界：只授權 approval_executor；write/admin 仍需 Gate 5 短效 approval key。
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+WITH agent_body AS (
+    SELECT jsonb_build_object(
+        'schema_version', 'awooop_agent_contract_v1',
+        'agent_id', 'approval_executor',
+        'display_name', 'Approval Executor',
+        'project_id', 'awoooi',
+        'purpose', 'Approved SSH execution through AwoooP MCP Gateway',
+        'allowed_scopes', jsonb_build_array('read', 'write', 'admin'),
+        'requires_gate5_for_scopes', jsonb_build_array('write', 'admin'),
+        'stage', 't9_ssh_approval_gateway'
+    ) AS body_json
+),
+inserted_revision AS (
+    INSERT INTO awooop_contract_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        version_major,
+        version_minor,
+        lifecycle_status,
+        body_json,
+        body_hash,
+        body_schema_version,
+        publisher_id,
+        published_at
+    )
+    SELECT
+        'awoooi',
+        'agent',
+        'approval_executor',
+        1,
+        0,
+        'active',
+        body_json,
+        encode(digest(body_json::text, 'sha256'), 'hex'),
+        'v1.0',
+        'migration:t9_ssh_approval_gateway',
+        NOW()
+    FROM agent_body
+    ON CONFLICT (project_id, contract_family, contract_id, version_major, version_minor)
+    DO NOTHING
+    RETURNING revision_id, project_id, contract_family, contract_id
+),
+chosen_revision AS (
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM inserted_revision
+    UNION ALL
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM awooop_contract_revisions
+    WHERE project_id = 'awoooi'
+      AND contract_family = 'agent'
+      AND contract_id = 'approval_executor'
+      AND version_major = 1
+      AND version_minor = 0
+      AND lifecycle_status = 'active'
+),
+upsert_pointer AS (
+    INSERT INTO awooop_active_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        active_revision_id,
+        updated_at
+    )
+    SELECT DISTINCT ON (project_id, contract_family, contract_id)
+        project_id,
+        contract_family,
+        contract_id,
+        revision_id,
+        NOW()
+    FROM chosen_revision
+    ORDER BY project_id, contract_family, contract_id, revision_id
+    ON CONFLICT (project_id, contract_family, contract_id)
+    DO UPDATE SET
+        active_revision_id = EXCLUDED.active_revision_id,
+        updated_at = NOW()
+    RETURNING contract_id
+)
+SELECT 'approval_executor_active_contracts', count(*) FROM upsert_pointer;
+
+WITH gateway_tools(tool_name, description, required_scope) AS (
+    VALUES
+        ('ssh_diagnose', 'SSH host diagnosis read', 'read'),
+        ('ssh_docker_restart', 'Approved Docker container restart over SSH', 'write'),
+        ('ssh_docker_compose_restart', 'Approved Docker Compose service restart over SSH', 'write'),
+        ('ssh_systemctl_restart', 'Approved systemd service restart over SSH', 'write'),
+        ('ssh_clear_docker_logs', 'Approved Docker log truncation over SSH', 'write'),
+        ('ssh_renew_ssl', 'Approved certbot renewal over SSH', 'write'),
+        ('ssh_reload_nginx', 'Approved nginx config test and reload over SSH', 'write'),
+        ('ssh_docker_prune', 'Approved Docker prune over SSH with provider disk guard', 'admin')
+),
+upsert_tools AS (
+    INSERT INTO awooop_mcp_tool_registry (
+        project_id,
+        tool_name,
+        tool_type,
+        description,
+        allowed_scopes,
+        environment_tags,
+        is_active,
+        updated_at
+    )
+    SELECT
+        'awoooi',
+        tool_name,
+        'mcp_server',
+        description,
+        jsonb_build_array(required_scope),
+        '{"env": "prod"}'::jsonb,
+        TRUE,
+        NOW()
+    FROM gateway_tools
+    ON CONFLICT (project_id, tool_name)
+    DO UPDATE SET
+        description = EXCLUDED.description,
+        allowed_scopes = EXCLUDED.allowed_scopes,
+        environment_tags = EXCLUDED.environment_tags,
+        is_active = TRUE,
+        updated_at = NOW()
+    RETURNING tool_id, tool_name, allowed_scopes
+),
+upsert_grants AS (
+    INSERT INTO awooop_mcp_grants (
+        project_id,
+        agent_id,
+        tool_id,
+        granted_by,
+        granted_scopes,
+        expires_at,
+        is_revoked,
+        revoked_at,
+        revoked_by
+    )
+    SELECT
+        'awoooi',
+        'approval_executor',
+        tool_id,
+        'migration:t9_ssh_approval_gateway',
+        allowed_scopes,
+        NULL,
+        FALSE,
+        NULL,
+        NULL
+    FROM upsert_tools
+    ON CONFLICT (project_id, agent_id, tool_id)
+    DO UPDATE SET
+        granted_by = EXCLUDED.granted_by,
+        granted_scopes = EXCLUDED.granted_scopes,
+        expires_at = NULL,
+        is_revoked = FALSE,
+        revoked_at = NULL,
+        revoked_by = NULL
+    RETURNING grant_id
+)
+SELECT
+    'approval_executor_ssh_gateway',
+    (SELECT count(*) FROM upsert_tools) AS tool_rows,
+    (SELECT count(*) FROM upsert_grants) AS grant_rows;

apps/api/migrations/awooop_awoooi_mcp_approval_executor_ssh_gateway_2026-05-13_down.sql

@@ -0,0 +1,43 @@
+-- Rollback for T9 approved SSH execution MCP Gateway seed.
+-- Contract revisions are append-only; rollback revokes approval_executor grants
+-- and deactivates only the write/admin tools introduced here.
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+UPDATE awooop_mcp_grants
+SET
+    is_revoked = TRUE,
+    revoked_at = NOW(),
+    revoked_by = 'rollback:t9_ssh_approval_gateway'
+WHERE project_id = 'awoooi'
+  AND agent_id = 'approval_executor'
+  AND granted_by = 'migration:t9_ssh_approval_gateway'
+  AND is_revoked = FALSE;
+
+UPDATE awooop_mcp_tool_registry
+SET
+    is_active = FALSE,
+    updated_at = NOW()
+WHERE project_id = 'awoooi'
+  AND tool_name IN (
+      'ssh_docker_restart',
+      'ssh_docker_compose_restart',
+      'ssh_systemctl_restart',
+      'ssh_clear_docker_logs',
+      'ssh_renew_ssl',
+      'ssh_reload_nginx',
+      'ssh_docker_prune'
+  );
+
+DELETE FROM awooop_active_revisions
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id = 'approval_executor';
+
+UPDATE awooop_contract_revisions
+SET lifecycle_status = 'revoked'
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id = 'approval_executor'
+  AND publisher_id = 'migration:t9_ssh_approval_gateway'
+  AND lifecycle_status = 'active';

apps/api/migrations/awooop_awoooi_mcp_auto_repair_executor_read_gateway_2026-05-14.sql

@@ -0,0 +1,166 @@
+-- T23: auto-repair executor read-only MCP Gateway seed
+-- 目的：讓 YAML_RULE/PlayBook 的只讀 SSH 診斷步驟經過 AwoooP MCP Gateway。
+-- 邊界：只授權 read scope；write/admin SSH 工具仍必須走 approval_executor + Gate 5。
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+WITH agent_body AS (
+    SELECT jsonb_build_object(
+        'schema_version', 'awooop_agent_contract_v1',
+        'agent_id', 'auto_repair_executor',
+        'display_name', 'Auto Repair Executor',
+        'project_id', 'awoooi',
+        'purpose', 'Read-only auto-repair diagnostics through AwoooP MCP Gateway',
+        'allowed_scopes', jsonb_build_array('read'),
+        'forbidden_scopes', jsonb_build_array('write', 'admin'),
+        'stage', 't23_auto_repair_diagnostic_gateway'
+    ) AS body_json
+),
+inserted_revision AS (
+    INSERT INTO awooop_contract_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        version_major,
+        version_minor,
+        lifecycle_status,
+        body_json,
+        body_hash,
+        body_schema_version,
+        publisher_id,
+        published_at
+    )
+    SELECT
+        'awoooi',
+        'agent',
+        'auto_repair_executor',
+        1,
+        0,
+        'active',
+        body_json,
+        encode(digest(body_json::text, 'sha256'), 'hex'),
+        'v1.0',
+        'migration:t23_auto_repair_executor_read_gateway',
+        NOW()
+    FROM agent_body
+    ON CONFLICT (project_id, contract_family, contract_id, version_major, version_minor)
+    DO NOTHING
+    RETURNING revision_id, project_id, contract_family, contract_id
+),
+chosen_revision AS (
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM inserted_revision
+    UNION ALL
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM awooop_contract_revisions
+    WHERE project_id = 'awoooi'
+      AND contract_family = 'agent'
+      AND contract_id = 'auto_repair_executor'
+      AND version_major = 1
+      AND version_minor = 0
+      AND lifecycle_status = 'active'
+),
+upsert_pointer AS (
+    INSERT INTO awooop_active_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        active_revision_id,
+        updated_at
+    )
+    SELECT DISTINCT ON (project_id, contract_family, contract_id)
+        project_id,
+        contract_family,
+        contract_id,
+        revision_id,
+        NOW()
+    FROM chosen_revision
+    ORDER BY project_id, contract_family, contract_id, revision_id
+    ON CONFLICT (project_id, contract_family, contract_id)
+    DO UPDATE SET
+        active_revision_id = EXCLUDED.active_revision_id,
+        updated_at = NOW()
+    RETURNING contract_id
+)
+SELECT 'auto_repair_executor_active_contracts', count(*) FROM upsert_pointer;
+
+WITH read_tools(tool_name, description) AS (
+    VALUES
+        ('ssh_diagnose', 'SSH host/container diagnosis read'),
+        ('ssh_get_top_processes', 'SSH top processes read'),
+        ('ssh_get_disk_usage', 'SSH disk usage read'),
+        ('ssh_get_memory_info', 'SSH memory info read'),
+        ('ssh_get_container_logs', 'SSH container logs read'),
+        ('ssh_get_container_status', 'SSH container status read'),
+        ('ssh_get_service_status', 'SSH service status read'),
+        ('ssh_check_port', 'SSH port check read'),
+        ('ssh_get_nginx_error_log', 'SSH nginx error log read'),
+        ('ssh_get_swap_info', 'SSH swap info read')
+),
+upsert_tools AS (
+    INSERT INTO awooop_mcp_tool_registry (
+        project_id,
+        tool_name,
+        tool_type,
+        description,
+        allowed_scopes,
+        environment_tags,
+        is_active,
+        updated_at
+    )
+    SELECT
+        'awoooi',
+        tool_name,
+        'mcp_server',
+        description,
+        '["read"]'::jsonb,
+        '{"env": "prod"}'::jsonb,
+        TRUE,
+        NOW()
+    FROM read_tools
+    ON CONFLICT (project_id, tool_name)
+    DO UPDATE SET
+        description = EXCLUDED.description,
+        allowed_scopes = EXCLUDED.allowed_scopes,
+        environment_tags = EXCLUDED.environment_tags,
+        is_active = TRUE,
+        updated_at = NOW()
+    RETURNING tool_id, tool_name, allowed_scopes
+),
+upsert_grants AS (
+    INSERT INTO awooop_mcp_grants (
+        project_id,
+        agent_id,
+        tool_id,
+        granted_by,
+        granted_scopes,
+        expires_at,
+        is_revoked,
+        revoked_at,
+        revoked_by
+    )
+    SELECT
+        'awoooi',
+        'auto_repair_executor',
+        tool_id,
+        'migration:t23_auto_repair_executor_read_gateway',
+        allowed_scopes,
+        NULL,
+        FALSE,
+        NULL,
+        NULL
+    FROM upsert_tools
+    ON CONFLICT (project_id, agent_id, tool_id)
+    DO UPDATE SET
+        granted_by = EXCLUDED.granted_by,
+        granted_scopes = EXCLUDED.granted_scopes,
+        expires_at = NULL,
+        is_revoked = FALSE,
+        revoked_at = NULL,
+        revoked_by = NULL
+    RETURNING grant_id
+)
+SELECT
+    'auto_repair_executor_read_gateway',
+    (SELECT count(*) FROM upsert_tools) AS tool_rows,
+    (SELECT count(*) FROM upsert_grants) AS grant_rows;

apps/api/migrations/awooop_awoooi_mcp_auto_repair_executor_read_gateway_2026-05-14_down.sql

@@ -0,0 +1,24 @@
+-- Rollback T23 auto-repair executor read-only MCP Gateway grant.
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+UPDATE awooop_mcp_grants
+SET is_revoked = TRUE,
+    revoked_at = NOW(),
+    revoked_by = 'rollback:t23_auto_repair_executor_read_gateway'
+WHERE project_id = 'awoooi'
+  AND agent_id = 'auto_repair_executor'
+  AND granted_by = 'migration:t23_auto_repair_executor_read_gateway';
+
+DELETE FROM awooop_active_revisions
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id = 'auto_repair_executor';
+
+UPDATE awooop_contract_revisions
+SET lifecycle_status = 'retired'
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id = 'auto_repair_executor'
+  AND publisher_id = 'migration:t23_auto_repair_executor_read_gateway'
+  AND lifecycle_status = 'active';

apps/api/migrations/awooop_awoooi_mcp_gateway_project_shadow_2026-05-13.sql

@@ -0,0 +1,25 @@
+-- =============================================================================
+-- AwoooP / AWOOOI MCP Gateway Shadow Onboarding
+-- 2026-05-13 Codex + ogt
+--
+-- 背景：
+--   AWOOOI 已完成 read-only MCP tool registry / grants seed，但 project 本身仍停在
+--   legacy_awoooi_default，會被 MCP Gateway Gate 1 正確攔截。
+--
+-- 邊界：
+--   只把 AWOOOI 租戶升到 shadow，讓既有 Gate 1 生效。
+--   write/admin tool 仍未授權；自動修復/破壞性動作不因本 migration 開放。
+-- =============================================================================
+
+BEGIN;
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+UPDATE awooop_projects
+SET
+    migration_mode = 'shadow',
+    updated_at = NOW()
+WHERE project_id = 'awoooi'
+  AND migration_mode = 'legacy_awoooi_default';
+
+COMMIT;

apps/api/migrations/awooop_awoooi_mcp_gateway_project_shadow_2026-05-13_down.sql

@@ -0,0 +1,20 @@
+-- =============================================================================
+-- Rollback: AwoooP / AWOOOI MCP Gateway Shadow Onboarding
+-- 2026-05-13 Codex + ogt
+--
+-- 只回退仍停在 shadow 的 AWOOOI；若已由人工/後續 migration 推進到 canary/active，
+-- 不自動降級。
+-- =============================================================================
+
+BEGIN;
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+UPDATE awooop_projects
+SET
+    migration_mode = 'legacy_awoooi_default',
+    updated_at = NOW()
+WHERE project_id = 'awoooi'
+  AND migration_mode = 'shadow';
+
+COMMIT;

apps/api/migrations/awooop_awoooi_mcp_read_gateway_seed_v3_2026-05-13.sql

@@ -0,0 +1,211 @@
+-- T7: awoooi read-only MCP Gateway seed
+-- 目的：讓決策前感官 MCP 能通過 AwoooP Gateway Gate 2/3，產生 first-class audit。
+-- 邊界：只授權 read scope；不授權 restart/delete/scale/apply/rollback 等 write/admin 工具。
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+WITH agent_seed(agent_id, display_name) AS (
+    VALUES
+        ('pre_decision_investigator', 'Pre-decision Investigator'),
+        ('post_execution_verifier', 'Post-execution Verifier')
+),
+agent_body AS (
+    SELECT
+        agent_id,
+        jsonb_build_object(
+            'schema_version', 'awooop_agent_contract_v1',
+            'agent_id', agent_id,
+            'display_name', display_name,
+            'project_id', 'awoooi',
+            'purpose', 'Read-only MCP sensing through AwoooP Gateway',
+            'allowed_scopes', jsonb_build_array('read'),
+            'forbidden_scopes', jsonb_build_array('write', 'admin'),
+            'stage', 't7_mcp_gateway_read_sense'
+        ) AS body_json
+    FROM agent_seed
+),
+inserted_revision AS (
+    INSERT INTO awooop_contract_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        version_major,
+        version_minor,
+        lifecycle_status,
+        body_json,
+        body_hash,
+        body_schema_version,
+        publisher_id,
+        published_at
+    )
+    SELECT
+        'awoooi',
+        'agent',
+        agent_id,
+        1,
+        0,
+        'active',
+        body_json,
+        encode(digest(body_json::text, 'sha256'), 'hex'),
+        'v1.0',
+        'migration:t7_mcp_gateway_read_seed',
+        NOW()
+    FROM agent_body
+    ON CONFLICT (project_id, contract_family, contract_id, version_major, version_minor)
+    DO NOTHING
+    RETURNING revision_id, project_id, contract_family, contract_id
+),
+chosen_revision AS (
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM inserted_revision
+    UNION ALL
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM awooop_contract_revisions
+    WHERE project_id = 'awoooi'
+      AND contract_family = 'agent'
+      AND contract_id IN (SELECT agent_id FROM agent_seed)
+      AND version_major = 1
+      AND version_minor = 0
+      AND lifecycle_status = 'active'
+),
+upsert_pointer AS (
+    INSERT INTO awooop_active_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        active_revision_id,
+        updated_at
+    )
+    SELECT DISTINCT ON (project_id, contract_family, contract_id)
+        project_id,
+        contract_family,
+        contract_id,
+        revision_id,
+        NOW()
+    FROM chosen_revision
+    ORDER BY project_id, contract_family, contract_id, revision_id
+    ON CONFLICT (project_id, contract_family, contract_id)
+    DO UPDATE SET
+        active_revision_id = EXCLUDED.active_revision_id,
+        updated_at = NOW()
+    RETURNING contract_id
+)
+SELECT 'active_agent_contracts', count(*) FROM upsert_pointer;
+
+WITH read_tools(tool_name, description) AS (
+    VALUES
+        ('k8s_get_pod_logs', 'Kubernetes pod logs read'),
+        ('k8s_get_events', 'Kubernetes events read'),
+        ('k8s_describe_pod', 'Kubernetes pod describe read'),
+        ('k8s_get_hpa_status', 'Kubernetes HPA status read'),
+        ('k8s_get_node_conditions', 'Kubernetes node conditions read'),
+        ('ssh_diagnose', 'SSH host diagnosis read'),
+        ('ssh_get_top_processes', 'SSH top processes read'),
+        ('ssh_get_disk_usage', 'SSH disk usage read'),
+        ('ssh_get_memory_info', 'SSH memory info read'),
+        ('ssh_get_container_logs', 'SSH container logs read'),
+        ('ssh_get_container_status', 'SSH container status read'),
+        ('ssh_get_service_status', 'SSH service status read'),
+        ('ssh_check_port', 'SSH port check read'),
+        ('ssh_get_nginx_error_log', 'SSH nginx error log read'),
+        ('ssh_get_swap_info', 'SSH swap info read'),
+        ('prometheus_query', 'Prometheus instant query read'),
+        ('prometheus_query_range', 'Prometheus range query read'),
+        ('prometheus_get_alert_history', 'Prometheus alert history read'),
+        ('gold_metrics', 'SigNoz gold metrics read'),
+        ('trace_url', 'SigNoz trace URL read'),
+        ('system_metrics', 'SigNoz system metrics read'),
+        ('query_logs', 'SigNoz logs read'),
+        ('error_logs_summary', 'SigNoz error logs summary read'),
+        ('list_approvals', 'Approval records read'),
+        ('get_approval', 'Approval detail read'),
+        ('list_incidents', 'Incident records read'),
+        ('list_timeline', 'Timeline records read'),
+        ('read_file', 'Filesystem allowlisted file read'),
+        ('list_directory', 'Filesystem allowlisted directory read'),
+        ('search_in_file', 'Filesystem allowlisted file search'),
+        ('list_dashboards', 'Grafana dashboards read'),
+        ('get_dashboard', 'Grafana dashboard read'),
+        ('get_panel_data', 'Grafana panel data read'),
+        ('generate_dashboard_url', 'Grafana dashboard URL read'),
+        ('search_runbook', 'Runbook semantic search read'),
+        ('get_index_stats', 'Runbook index stats read'),
+        ('argocd_list_apps', 'ArgoCD apps read'),
+        ('argocd_get_app_status', 'ArgoCD app status read'),
+        ('argocd_get_sync_history', 'ArgoCD sync history read'),
+        ('sentry_list_issues', 'Sentry issues read'),
+        ('sentry_get_issue', 'Sentry issue detail read'),
+        ('sentry_search_issues', 'Sentry issue search read')
+),
+upsert_tools AS (
+    INSERT INTO awooop_mcp_tool_registry (
+        project_id,
+        tool_name,
+        tool_type,
+        description,
+        allowed_scopes,
+        environment_tags,
+        is_active,
+        updated_at
+    )
+    SELECT
+        'awoooi',
+        tool_name,
+        'mcp_server',
+        description,
+        '["read"]'::jsonb,
+        '{"env": "prod"}'::jsonb,
+        TRUE,
+        NOW()
+    FROM read_tools
+    ON CONFLICT (project_id, tool_name)
+    DO UPDATE SET
+        description = EXCLUDED.description,
+        allowed_scopes = EXCLUDED.allowed_scopes,
+        environment_tags = EXCLUDED.environment_tags,
+        is_active = TRUE,
+        updated_at = NOW()
+    RETURNING tool_id
+),
+grant_agents(agent_id) AS (
+    VALUES
+        ('pre_decision_investigator'),
+        ('post_execution_verifier')
+),
+upsert_grants AS (
+    INSERT INTO awooop_mcp_grants (
+        project_id,
+        agent_id,
+        tool_id,
+        granted_by,
+        granted_scopes,
+        expires_at,
+        is_revoked,
+        revoked_at,
+        revoked_by
+    )
+    SELECT
+        'awoooi',
+        grant_agents.agent_id,
+        upsert_tools.tool_id,
+        'migration:t7_mcp_gateway_read_seed',
+        '["read"]'::jsonb,
+        NULL,
+        FALSE,
+        NULL,
+        NULL
+    FROM upsert_tools
+    CROSS JOIN grant_agents
+    ON CONFLICT (project_id, agent_id, tool_id)
+    DO UPDATE SET
+        granted_scopes = EXCLUDED.granted_scopes,
+        expires_at = NULL,
+        is_revoked = FALSE,
+        revoked_at = NULL,
+        revoked_by = NULL
+    RETURNING grant_id
+)
+SELECT
+    'awoooi_read_tools',
+    (SELECT count(*) FROM upsert_tools) AS tool_rows,
+    (SELECT count(*) FROM upsert_grants) AS grant_rows;

apps/api/migrations/awooop_awoooi_mcp_read_gateway_seed_v3_2026-05-13_down.sql

@@ -0,0 +1,77 @@
+-- Rollback for T7 awoooi read-only MCP Gateway seed.
+-- Contract revisions are append-only; rollback revokes grants and deactivates the seeded read tools.
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+UPDATE awooop_mcp_grants
+SET
+    is_revoked = TRUE,
+    revoked_at = NOW(),
+    revoked_by = 'rollback:t7_mcp_gateway_read_seed'
+WHERE project_id = 'awoooi'
+  AND agent_id IN ('pre_decision_investigator', 'post_execution_verifier')
+  AND granted_by = 'migration:t7_mcp_gateway_read_seed'
+  AND is_revoked = FALSE;
+
+UPDATE awooop_mcp_tool_registry
+SET
+    is_active = FALSE,
+    updated_at = NOW()
+WHERE project_id = 'awoooi'
+  AND tool_name IN (
+      'k8s_get_pod_logs',
+      'k8s_get_events',
+      'k8s_describe_pod',
+      'k8s_get_hpa_status',
+      'k8s_get_node_conditions',
+      'ssh_diagnose',
+      'ssh_get_top_processes',
+      'ssh_get_disk_usage',
+      'ssh_get_memory_info',
+      'ssh_get_container_logs',
+      'ssh_get_container_status',
+      'ssh_get_service_status',
+      'ssh_check_port',
+      'ssh_get_nginx_error_log',
+      'ssh_get_swap_info',
+      'prometheus_query',
+      'prometheus_query_range',
+      'prometheus_get_alert_history',
+      'gold_metrics',
+      'trace_url',
+      'system_metrics',
+      'query_logs',
+      'error_logs_summary',
+      'list_approvals',
+      'get_approval',
+      'list_incidents',
+      'list_timeline',
+      'read_file',
+      'list_directory',
+      'search_in_file',
+      'list_dashboards',
+      'get_dashboard',
+      'get_panel_data',
+      'generate_dashboard_url',
+      'search_runbook',
+      'get_index_stats',
+      'argocd_list_apps',
+      'argocd_get_app_status',
+      'argocd_get_sync_history',
+      'sentry_list_issues',
+      'sentry_get_issue',
+      'sentry_search_issues'
+  );
+
+DELETE FROM awooop_active_revisions
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id IN ('pre_decision_investigator', 'post_execution_verifier');
+
+UPDATE awooop_contract_revisions
+SET lifecycle_status = 'revoked'
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id IN ('pre_decision_investigator', 'post_execution_verifier')
+  AND publisher_id = 'migration:t7_mcp_gateway_read_seed'
+  AND lifecycle_status = 'active';

apps/api/migrations/awooop_awoooi_mcp_read_gateway_seed_v4_2026-05-13.sql

@@ -0,0 +1,213 @@
+-- T7: awoooi read-only MCP Gateway seed
+-- 目的：讓決策前感官 MCP 能通過 AwoooP Gateway Gate 2/3，產生 first-class audit。
+-- 邊界：只授權 read scope；不授權 restart/delete/scale/apply/rollback 等 write/admin 工具。
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+WITH agent_seed(agent_id, display_name) AS (
+    VALUES
+        ('pre_decision_investigator', 'Pre-decision Investigator'),
+        ('post_execution_verifier', 'Post-execution Verifier')
+),
+agent_body AS (
+    SELECT
+        agent_id,
+        jsonb_build_object(
+            'schema_version', 'awooop_agent_contract_v1',
+            'agent_id', agent_id,
+            'display_name', display_name,
+            'project_id', 'awoooi',
+            'purpose', 'Read-only MCP sensing through AwoooP Gateway',
+            'allowed_scopes', jsonb_build_array('read'),
+            'forbidden_scopes', jsonb_build_array('write', 'admin'),
+            'stage', 't7_mcp_gateway_read_sense'
+        ) AS body_json
+    FROM agent_seed
+),
+inserted_revision AS (
+    INSERT INTO awooop_contract_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        version_major,
+        version_minor,
+        lifecycle_status,
+        body_json,
+        body_hash,
+        body_schema_version,
+        publisher_id,
+        published_at
+    )
+    SELECT
+        'awoooi',
+        'agent',
+        agent_id,
+        1,
+        0,
+        'active',
+        body_json,
+        encode(digest(body_json::text, 'sha256'), 'hex'),
+        'v1.0',
+        'migration:t7_mcp_gateway_read_seed',
+        NOW()
+    FROM agent_body
+    ON CONFLICT (project_id, contract_family, contract_id, version_major, version_minor)
+    DO NOTHING
+    RETURNING revision_id, project_id, contract_family, contract_id
+),
+chosen_revision AS (
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM inserted_revision
+    UNION ALL
+    SELECT revision_id, project_id, contract_family, contract_id
+    FROM awooop_contract_revisions
+    WHERE project_id = 'awoooi'
+      AND contract_family = 'agent'
+      AND contract_id IN (SELECT agent_id FROM agent_seed)
+      AND version_major = 1
+      AND version_minor = 0
+      AND lifecycle_status = 'active'
+),
+upsert_pointer AS (
+    INSERT INTO awooop_active_revisions (
+        project_id,
+        contract_family,
+        contract_id,
+        active_revision_id,
+        updated_at
+    )
+    SELECT DISTINCT ON (project_id, contract_family, contract_id)
+        project_id,
+        contract_family,
+        contract_id,
+        revision_id,
+        NOW()
+    FROM chosen_revision
+    ORDER BY project_id, contract_family, contract_id, revision_id
+    ON CONFLICT (project_id, contract_family, contract_id)
+    DO UPDATE SET
+        active_revision_id = EXCLUDED.active_revision_id,
+        updated_at = NOW()
+    RETURNING contract_id
+)
+SELECT 'active_agent_contracts', count(*) FROM upsert_pointer;
+
+WITH read_tools(tool_name, description) AS (
+    VALUES
+        ('k8s_get_pod_logs', 'Kubernetes pod logs read'),
+        ('k8s_get_events', 'Kubernetes events read'),
+        ('k8s_describe_pod', 'Kubernetes pod describe read'),
+        ('k8s_get_hpa_status', 'Kubernetes HPA status read'),
+        ('k8s_get_node_conditions', 'Kubernetes node conditions read'),
+        ('ssh_diagnose', 'SSH host diagnosis read'),
+        ('ssh_get_top_processes', 'SSH top processes read'),
+        ('ssh_get_disk_usage', 'SSH disk usage read'),
+        ('ssh_get_memory_info', 'SSH memory info read'),
+        ('ssh_get_container_logs', 'SSH container logs read'),
+        ('ssh_get_container_status', 'SSH container status read'),
+        ('ssh_get_service_status', 'SSH service status read'),
+        ('ssh_check_port', 'SSH port check read'),
+        ('ssh_get_nginx_error_log', 'SSH nginx error log read'),
+        ('ssh_get_swap_info', 'SSH swap info read'),
+        ('prometheus_query', 'Prometheus instant query read'),
+        ('prometheus_query_range', 'Prometheus range query read'),
+        ('prometheus_get_alert_history', 'Prometheus alert history read'),
+        ('gold_metrics', 'SigNoz gold metrics read'),
+        ('trace_url', 'SigNoz trace URL read'),
+        ('system_metrics', 'SigNoz system metrics read'),
+        ('query_logs', 'SigNoz logs read'),
+        ('error_logs_summary', 'SigNoz error logs summary read'),
+        ('list_approvals', 'Approval records read'),
+        ('get_approval', 'Approval detail read'),
+        ('list_incidents', 'Incident records read'),
+        ('list_timeline', 'Timeline records read'),
+        ('read_file', 'Filesystem allowlisted file read'),
+        ('list_directory', 'Filesystem allowlisted directory read'),
+        ('search_in_file', 'Filesystem allowlisted file search'),
+        ('list_dashboards', 'Grafana dashboards read'),
+        ('get_dashboard', 'Grafana dashboard read'),
+        ('get_panel_data', 'Grafana panel data read'),
+        ('generate_dashboard_url', 'Grafana dashboard URL read'),
+        ('search_runbook', 'Runbook semantic search read'),
+        ('get_index_stats', 'Runbook index stats read'),
+        ('argocd_list_apps', 'ArgoCD apps read'),
+        ('argocd_get_app_status', 'ArgoCD app status read'),
+        ('argocd_get_sync_history', 'ArgoCD sync history read'),
+        ('sentry_list_issues', 'Sentry issues read'),
+        ('sentry_get_issue', 'Sentry issue detail read'),
+        ('sentry_search_issues', 'Sentry issue search read')
+),
+upsert_tools AS (
+    INSERT INTO awooop_mcp_tool_registry (
+        project_id,
+        tool_name,
+        tool_type,
+        description,
+        allowed_scopes,
+        environment_tags,
+        is_active,
+        updated_at
+    )
+    SELECT
+        'awoooi',
+        tool_name,
+        'mcp_server',
+        description,
+        '["read"]'::jsonb,
+        '{"env": "prod"}'::jsonb,
+        TRUE,
+        NOW()
+    FROM read_tools
+    ON CONFLICT (project_id, tool_name)
+    DO UPDATE SET
+        description = EXCLUDED.description,
+        allowed_scopes = EXCLUDED.allowed_scopes,
+        environment_tags = EXCLUDED.environment_tags,
+        is_active = TRUE,
+        updated_at = NOW()
+    RETURNING tool_id
+),
+grant_agents(agent_id) AS (
+    VALUES
+        ('pre_decision_investigator'),
+        ('post_execution_verifier')
+),
+upsert_grants AS (
+    INSERT INTO awooop_mcp_grants (
+        project_id,
+        agent_id,
+        tool_id,
+        granted_by,
+        granted_scopes,
+        expires_at,
+        is_revoked,
+        revoked_at,
+        revoked_by
+    )
+    SELECT
+        'awoooi',
+        grant_agents.agent_id,
+        upsert_tools.tool_id,
+        'migration:t7_mcp_gateway_read_seed',
+        '["read"]'::jsonb,
+        NULL,
+        FALSE,
+        NULL,
+        NULL
+    FROM upsert_tools
+    CROSS JOIN grant_agents
+    ON CONFLICT (project_id, agent_id, tool_id)
+    DO UPDATE SET
+        granted_scopes = EXCLUDED.granted_scopes,
+        expires_at = NULL,
+        is_revoked = FALSE,
+        revoked_at = NULL,
+        revoked_by = NULL
+    RETURNING grant_id
+)
+SELECT
+    'awoooi_read_tools',
+    (SELECT count(*) FROM upsert_tools) AS tool_rows,
+    (SELECT count(*) FROM upsert_grants) AS grant_rows;
+
+-- v4 exists only to retrigger run-migration after Gitea skipped the v2->v3 rename-only push.

apps/api/migrations/awooop_awoooi_mcp_read_gateway_seed_v4_2026-05-13_down.sql

@@ -0,0 +1,79 @@
+-- Rollback for T7 awoooi read-only MCP Gateway seed.
+-- Contract revisions are append-only; rollback revokes grants and deactivates the seeded read tools.
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+UPDATE awooop_mcp_grants
+SET
+    is_revoked = TRUE,
+    revoked_at = NOW(),
+    revoked_by = 'rollback:t7_mcp_gateway_read_seed'
+WHERE project_id = 'awoooi'
+  AND agent_id IN ('pre_decision_investigator', 'post_execution_verifier')
+  AND granted_by = 'migration:t7_mcp_gateway_read_seed'
+  AND is_revoked = FALSE;
+
+UPDATE awooop_mcp_tool_registry
+SET
+    is_active = FALSE,
+    updated_at = NOW()
+WHERE project_id = 'awoooi'
+  AND tool_name IN (
+      'k8s_get_pod_logs',
+      'k8s_get_events',
+      'k8s_describe_pod',
+      'k8s_get_hpa_status',
+      'k8s_get_node_conditions',
+      'ssh_diagnose',
+      'ssh_get_top_processes',
+      'ssh_get_disk_usage',
+      'ssh_get_memory_info',
+      'ssh_get_container_logs',
+      'ssh_get_container_status',
+      'ssh_get_service_status',
+      'ssh_check_port',
+      'ssh_get_nginx_error_log',
+      'ssh_get_swap_info',
+      'prometheus_query',
+      'prometheus_query_range',
+      'prometheus_get_alert_history',
+      'gold_metrics',
+      'trace_url',
+      'system_metrics',
+      'query_logs',
+      'error_logs_summary',
+      'list_approvals',
+      'get_approval',
+      'list_incidents',
+      'list_timeline',
+      'read_file',
+      'list_directory',
+      'search_in_file',
+      'list_dashboards',
+      'get_dashboard',
+      'get_panel_data',
+      'generate_dashboard_url',
+      'search_runbook',
+      'get_index_stats',
+      'argocd_list_apps',
+      'argocd_get_app_status',
+      'argocd_get_sync_history',
+      'sentry_list_issues',
+      'sentry_get_issue',
+      'sentry_search_issues'
+  );
+
+DELETE FROM awooop_active_revisions
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id IN ('pre_decision_investigator', 'post_execution_verifier');
+
+UPDATE awooop_contract_revisions
+SET lifecycle_status = 'revoked'
+WHERE project_id = 'awoooi'
+  AND contract_family = 'agent'
+  AND contract_id IN ('pre_decision_investigator', 'post_execution_verifier')
+  AND publisher_id = 'migration:t7_mcp_gateway_read_seed'
+  AND lifecycle_status = 'active';
+
+-- v4 rollback companion for the retrigger migration.

apps/api/migrations/awooop_awoooi_mcp_rollout_verifier_seed_2026-05-13.sql

@@ -0,0 +1,77 @@
+-- T16 verifier gap: allow rollout status evidence through AwoooP MCP Gateway.
+-- Boundary: read-only scope only; no restart/delete/scale grant is added here.
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+WITH upsert_tool AS (
+    INSERT INTO awooop_mcp_tool_registry (
+        project_id,
+        tool_name,
+        tool_type,
+        description,
+        allowed_scopes,
+        environment_tags,
+        is_active,
+        updated_at
+    )
+    VALUES (
+        'awoooi',
+        'k8s_watch_rollout',
+        'mcp_server',
+        'Kubernetes deployment rollout status read',
+        '["read"]'::jsonb,
+        '{"env": "prod"}'::jsonb,
+        TRUE,
+        NOW()
+    )
+    ON CONFLICT (project_id, tool_name)
+    DO UPDATE SET
+        description = EXCLUDED.description,
+        allowed_scopes = EXCLUDED.allowed_scopes,
+        environment_tags = EXCLUDED.environment_tags,
+        is_active = TRUE,
+        updated_at = NOW()
+    RETURNING tool_id
+),
+grant_agents(agent_id) AS (
+    VALUES
+        ('pre_decision_investigator'),
+        ('post_execution_verifier')
+),
+upsert_grants AS (
+    INSERT INTO awooop_mcp_grants (
+        project_id,
+        agent_id,
+        tool_id,
+        granted_by,
+        granted_scopes,
+        expires_at,
+        is_revoked,
+        revoked_at,
+        revoked_by
+    )
+    SELECT
+        'awoooi',
+        grant_agents.agent_id,
+        upsert_tool.tool_id,
+        'migration:t16_rollout_verifier_seed',
+        '["read"]'::jsonb,
+        NULL,
+        FALSE,
+        NULL,
+        NULL
+    FROM upsert_tool
+    CROSS JOIN grant_agents
+    ON CONFLICT (project_id, agent_id, tool_id)
+    DO UPDATE SET
+        granted_scopes = EXCLUDED.granted_scopes,
+        expires_at = NULL,
+        is_revoked = FALSE,
+        revoked_at = NULL,
+        revoked_by = NULL
+    RETURNING grant_id
+)
+SELECT
+    'k8s_watch_rollout_read_grants' AS seed,
+    (SELECT count(*) FROM upsert_tool) AS tool_rows,
+    (SELECT count(*) FROM upsert_grants) AS grant_rows;

apps/api/migrations/awooop_awoooi_mcp_rollout_verifier_seed_2026-05-13_down.sql

@@ -0,0 +1,24 @@
+-- Roll back T16 rollout verifier read grant seed.
+
+SELECT set_config('app.project_id', 'awoooi', FALSE);
+
+UPDATE awooop_mcp_grants
+SET
+    is_revoked = TRUE,
+    revoked_at = NOW(),
+    revoked_by = 'migration:t16_rollout_verifier_seed_down'
+WHERE project_id = 'awoooi'
+  AND agent_id IN ('pre_decision_investigator', 'post_execution_verifier')
+  AND tool_id IN (
+      SELECT tool_id
+      FROM awooop_mcp_tool_registry
+      WHERE project_id = 'awoooi'
+        AND tool_name = 'k8s_watch_rollout'
+  );
+
+UPDATE awooop_mcp_tool_registry
+SET
+    is_active = FALSE,
+    updated_at = NOW()
+WHERE project_id = 'awoooi'
+  AND tool_name = 'k8s_watch_rollout';

apps/api/migrations/awooop_phase7_inbound_truth_chain_columns_2026-05-13.sql

@@ -0,0 +1,21 @@
+-- AwoooP Phase 7 T15b: inbound event truth-chain columns
+--
+-- Purpose:
+--   Telegram cards are only the notification surface. Operators need a
+--   redacted replay envelope for inbound alerts so Alertmanager, Sentry, and
+--   SignOz events can be correlated with incidents, approvals, logs, and
+--   automation decisions without storing raw secrets or PII.
+
+ALTER TABLE awooop_conversation_event
+    ADD COLUMN IF NOT EXISTS content_redacted TEXT,
+    ADD COLUMN IF NOT EXISTS redaction_version VARCHAR(32) NOT NULL DEFAULT 'audit_sink_v1',
+    ADD COLUMN IF NOT EXISTS source_envelope JSONB NOT NULL DEFAULT '{}'::jsonb;
+
+COMMENT ON COLUMN awooop_conversation_event.content_redacted IS
+    'Full inbound event content after audit_sink redaction; raw unredacted payload text is not stored.';
+
+COMMENT ON COLUMN awooop_conversation_event.redaction_version IS
+    'Redaction algorithm/version used for content_redacted and source_envelope.';
+
+COMMENT ON COLUMN awooop_conversation_event.source_envelope IS
+    'Redacted source metadata for inbound replay/audit, including payload hash, provider, source refs, and log correlation hints.';

apps/api/migrations/awooop_phase7_inbound_truth_chain_columns_2026-05-13_down.sql

@@ -0,0 +1,6 @@
+-- Rollback for AwoooP Phase 7 T15b inbound truth-chain columns.
+-- Safe only if no consumers depend on the redacted replay fields.
+
+ALTER TABLE awooop_conversation_event DROP COLUMN IF EXISTS source_envelope;
+ALTER TABLE awooop_conversation_event DROP COLUMN IF EXISTS redaction_version;
+ALTER TABLE awooop_conversation_event DROP COLUMN IF EXISTS content_redacted;

apps/api/migrations/awooop_phase7_outbound_truth_chain_columns_2026-05-12.sql

@@ -0,0 +1,21 @@
+-- AwoooP Phase 7 T1: outbound message truth-chain columns
+--
+-- Purpose:
+--   Telegram must remain a summary channel, but the operator console needs a
+--   complete redacted replay of the rendered card and the source envelope that
+--   produced it. Store redacted content only; raw unredacted Telegram text stays
+--   out of PostgreSQL.
+
+ALTER TABLE awooop_outbound_message
+    ADD COLUMN IF NOT EXISTS content_redacted TEXT,
+    ADD COLUMN IF NOT EXISTS redaction_version VARCHAR(32) NOT NULL DEFAULT 'audit_sink_v1',
+    ADD COLUMN IF NOT EXISTS source_envelope JSONB NOT NULL DEFAULT '{}'::jsonb;
+
+COMMENT ON COLUMN awooop_outbound_message.content_redacted IS
+    'Full rendered outbound content after audit_sink redaction; raw unredacted text is not stored.';
+
+COMMENT ON COLUMN awooop_outbound_message.redaction_version IS
+    'Redaction algorithm/version used for content_redacted and source_envelope.';
+
+COMMENT ON COLUMN awooop_outbound_message.source_envelope IS
+    'Redacted source metadata for replay/audit, including payload hash and adapter context.';

apps/api/migrations/awooop_phase7_outbound_truth_chain_columns_2026-05-12_down.sql

@@ -0,0 +1,6 @@
+-- Rollback for AwoooP Phase 7 T1 outbound truth-chain columns.
+-- Safe only if no consumers depend on the redacted replay fields.
+
+ALTER TABLE awooop_outbound_message DROP COLUMN IF EXISTS source_envelope;
+ALTER TABLE awooop_outbound_message DROP COLUMN IF EXISTS redaction_version;
+ALTER TABLE awooop_outbound_message DROP COLUMN IF EXISTS content_redacted;

apps/api/pyproject.toml

@@ -46,6 +46,10 @@ dependencies = [
     # 2026-04-16 ogt + Claude Sonnet 4.6: SSH MCP sensor 修復 — asyncssh 缺失導致 sensors_succeeded=0
     # 根因: ssh_provider.py 中 import asyncssh 在 try/except 外，所有 15 個 SSH tool 直接 ImportError
     "asyncssh>=2.14.0",
+    # 2026-05-31 Codex: AwoooP truth-chain Ansible runtime gate 需要
+    # production API image 內真的存在 ansible-playbook，否則只能顯示
+    # candidate audit，無法進入 check-mode executor readiness。
+    "ansible-core>=2.16.0,<2.18.0",
 ]
 
 # [tool.uv.sources]

apps/api/requirements.txt

@@ -58,3 +58,8 @@ pytest>=7.4.0
 pytest-asyncio>=0.23.0
 ruff>=0.1.0
 sentry-sdk[fastapi]>=2.0.0
+
+# AwoooP Ansible runtime readiness
+# 2026-05-31 Codex: production API image must include ansible-playbook before
+# truth-chain can honestly mark check-mode executor readiness as available.
+ansible-core>=2.16.0,<2.18.0

apps/api/scripts/awooop_phase1_batch1_backfill.py

@@ -9,7 +9,7 @@ AwoooP Phase 1 Batch 1 回填腳本
   awooop_phase1_batch1_rls_2026-05-04.sql Step A（ADD COLUMN nullable）已執行
 
 執行方式：
-  export DATABASE_URL="postgresql+asyncpg://awoooi:<password>@192.168.0.188:5432/awoooi_prod"
+  從 secret manager / operator vault 設定 DATABASE_URL，禁止在指令或檔案中寫入 URL。
   cd apps/api && python scripts/awooop_phase1_batch1_backfill.py
 
 2026-05-04 ogt + Claude Sonnet 4.6（ADR-118 Batch 1 C-3 修正）

apps/api/scripts/reembed_bge_m3.py

@@ -37,6 +37,7 @@ logging = structlog.get_logger(__name__)
 OLLAMA_URL = os.getenv("OLLAMA_URL", "http://34.143.170.20:11434")
 EMBEDDING_MODEL = "bge-m3:latest"
 EXPECTED_DIM = 1024
+PROJECT_ID = os.getenv("AWOOOP_PROJECT_ID", "awoooi")
 
 
 async def embed_text(client: httpx.AsyncClient, text: str) -> list[float]:
@@ -162,6 +163,7 @@ async def main(dry_run: bool, batch_size: int) -> None:
 
         conn = await asyncpg.connect(database_url)
         try:
+            await conn.execute("SELECT set_config('app.project_id', $1, FALSE)", PROJECT_ID)
             # 統計待嵌入筆數
             rag_null = await conn.fetchval("SELECT COUNT(*) FROM rag_chunks WHERE embedding IS NULL")
             pb_null = await conn.fetchval("SELECT COUNT(*) FROM playbook_embeddings WHERE embedding IS NULL")

apps/api/scripts/run_migration.py

@@ -15,7 +15,7 @@ from sqlalchemy import text
 from sqlalchemy.ext.asyncio import create_async_engine
 
 # 2026-04-22 ogt: 移除硬碼 changeme，改為讀取環境變數（強制要求設定）。
-# 執行前: export DATABASE_URL="postgresql+asyncpg://awoooi:<password>@192.168.0.188:5432/awoooi_prod"
+# 執行前: 從 secret manager / operator vault 設定 DATABASE_URL，禁止在指令或檔案中寫入 URL。
 DATABASE_URL = os.environ["DATABASE_URL"]
 
 MIGRATION_SQLS = [

apps/api/src/api/v1/ai_governance.py

@@ -22,17 +22,48 @@ from datetime import datetime
 from typing import Annotated
 
 import structlog
-from fastapi import APIRouter, Query
+from fastapi import APIRouter, HTTPException, Query
 
 from src.models.governance import (
     GovernanceEventsResponse,
     GovernanceQueueResponse,
     GovernanceSummaryResponse,
+    KnowledgeReviewDraftArchiveRequest,
+    KnowledgeReviewDraftArchiveResponse,
+    KnowledgeReviewDraftDedupeResponse,
+    KnowledgeStaleCandidatesResponse,
+    KnowledgeStaleOwnerReviewBatchQueueRequest,
+    KnowledgeStaleOwnerReviewBatchQueueResponse,
+    KnowledgeStaleOwnerReviewBurnDownResponse,
+    KnowledgeStaleOwnerReviewCompleteRequest,
+    KnowledgeStaleOwnerReviewCompleteResponse,
+    KnowledgeStaleOwnerReviewCompletionBatchPreviewRequest,
+    KnowledgeStaleOwnerReviewCompletionBatchPreviewResponse,
+    KnowledgeStaleOwnerReviewCompletionQueueResponse,
+    KnowledgeStaleOwnerReviewInboxResponse,
+    KnowledgeStaleOwnerReviewRequest,
+    KnowledgeStaleOwnerReviewResponse,
+)
+from src.services.governance_km_review_service import (
+    KmReviewDraftArchiveError,
+    archive_km_review_draft_duplicates,
+)
+from src.services.governance_km_stale_review_service import (
+    KmStaleOwnerReviewError,
+    batch_queue_km_stale_owner_reviews,
+    complete_km_stale_owner_review,
+    preview_km_stale_owner_review_completion_batch,
+    query_km_stale_owner_review_burndown,
+    query_km_stale_owner_review_completion_queue,
+    query_km_stale_owner_review_inbox,
+    queue_km_stale_owner_review,
 )
 from src.services.governance_query_service import (
     query_governance_events,
     query_governance_queue,
     query_governance_summary,
+    query_km_review_draft_dedupe,
+    query_km_stale_candidates,
 )
 
 logger = structlog.get_logger(__name__)
@@ -46,6 +77,7 @@ router = APIRouter()
 
 @router.get("/ai/governance/events", response_model=GovernanceEventsResponse)
 async def get_governance_events(
+    event_id: Annotated[list[str] | None, Query(alias="event_id")] = None,
     event_type: Annotated[list[str] | None, Query(alias="event_type")] = None,
     from_: Annotated[datetime | None, Query(alias="from")] = None,
     to: Annotated[datetime | None, Query(alias="to")] = None,
@@ -58,6 +90,7 @@ async def get_governance_events(
     查詢 AI 治理事件列表（分頁）。
 
     - event_type: 多值過濾（可重複傳）
+    - event_id: 多值精準過濾（可重複傳），供 Telegram 詳情 / 歷史與 Work Items 錨點回看
     - from / to: ISO 8601 時間範圍（URL 傳 from 參數）
     - status: resolved / unresolved
     - severity: critical / warning / info（由 event_type 映射決定）
@@ -66,6 +99,7 @@ async def get_governance_events(
     """
     logger.debug(
         "governance_events_request",
+        event_ids=event_id,
         event_types=event_type,
         from_=from_,
         to=to,
@@ -75,6 +109,7 @@ async def get_governance_events(
         size=size,
     )
     return await query_governance_events(
+        event_ids=event_id,
         event_types=event_type,
         from_dt=from_,
         to_dt=to,
@@ -93,8 +128,9 @@ async def get_governance_events(
 async def get_governance_queue(
     dispatch_status: Annotated[
         str,
-        Query(pattern="^(pending|dispatched|succeeded|failed)$"),
+        Query(pattern="^(all|pending|dispatched|executing|succeeded|failed|skipped|cancelled)$"),
     ] = "pending",
+    event_type: Annotated[list[str] | None, Query(alias="event_type")] = None,
     page: Annotated[int, Query(ge=1)] = 1,
     size: Annotated[int, Query(ge=10, le=100)] = 20,
 ) -> GovernanceQueueResponse:
@@ -104,22 +140,360 @@ async def get_governance_queue(
     governance_remediation_dispatch 表由 Track D 建立，尚未完成時
     本 endpoint 回傳 { table_pending: true, items: [], total: 0 }，不拋 500。
 
-    - dispatch_status: pending（default）/ dispatched / succeeded / failed
+    - dispatch_status: pending（default）/ dispatched / executing / succeeded / failed / skipped / cancelled / all
+    - event_type: 多值過濾（可重複傳）
     - page / size: 分頁
     """
     logger.debug(
         "governance_queue_request",
         dispatch_status=dispatch_status,
+        event_type=event_type,
         page=page,
         size=size,
     )
     return await query_governance_queue(
         dispatch_status=dispatch_status,
+        event_types=event_type,
         page=page,
         size=size,
     )
 
 
+# =============================================================================
+# GET /api/v1/ai/governance/km-review-drafts/dedupe
+# =============================================================================
+
+@router.get(
+    "/ai/governance/km-review-drafts/dedupe",
+    response_model=KnowledgeReviewDraftDedupeResponse,
+)
+async def get_km_review_draft_dedupe(
+    limit: Annotated[int, Query(ge=10, le=200)] = 100,
+) -> KnowledgeReviewDraftDedupeResponse:
+    """
+    查詢 Hermes KM healthcheck review drafts 的去重 read model。
+
+    這是 read-only owner review surface：只回傳 canonical / duplicate /
+    owner_action，不自動 archive、不自動 approve/publish KM。
+    """
+    logger.debug("km_review_draft_dedupe_request", limit=limit)
+    return await query_km_review_draft_dedupe(limit=limit)
+
+
+# =============================================================================
+# POST /api/v1/ai/governance/km-review-drafts/dedupe/{event_id}/archive-duplicates
+# =============================================================================
+
+@router.post(
+    "/ai/governance/km-review-drafts/dedupe/{governance_event_id}/archive-duplicates",
+    response_model=KnowledgeReviewDraftArchiveResponse,
+)
+async def post_km_review_draft_archive_duplicates(
+    governance_event_id: str,
+    request: KnowledgeReviewDraftArchiveRequest,
+) -> KnowledgeReviewDraftArchiveResponse:
+    """
+    Owner 審核後封存 Hermes KM healthcheck duplicate review drafts。
+
+    這不是 read endpoint：必須明確傳 owner_approved=true，且後端會重新比對
+    最新 dedupe plan。封存為 KnowledgeEntry.status=archived，不刪除資料。
+    """
+    logger.info(
+        "km_review_draft_archive_request",
+        governance_event_id=governance_event_id,
+        canonical_entry_id=request.canonical_entry_id,
+        duplicate_count=len(request.duplicate_entry_ids),
+        owner=request.owner,
+        dry_run=request.dry_run,
+        owner_approved=request.owner_approved,
+    )
+    try:
+        return await archive_km_review_draft_duplicates(
+            governance_event_id=governance_event_id,
+            request=request,
+        )
+    except KmReviewDraftArchiveError as exc:
+        raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
+
+
+# =============================================================================
+# GET /api/v1/ai/governance/km-stale-candidates
+# =============================================================================
+
+@router.get(
+    "/ai/governance/km-stale-candidates",
+    response_model=KnowledgeStaleCandidatesResponse,
+)
+async def get_km_stale_candidates(
+    project_id: Annotated[str, Query(min_length=1, max_length=64)] = "awoooi",
+    limit: Annotated[int, Query(ge=5, le=100)] = 20,
+) -> KnowledgeStaleCandidatesResponse:
+    """
+    查詢 stale KM 的 read-only 優先處理清單。
+
+    Hermes 可以用這個 read model 產生 KM 更新草稿；owner console 則能先看
+    哪些條目有 Incident / Sentry / SigNoz / PlayBook 脈絡，避免只看到總數。
+    """
+    logger.debug(
+        "km_stale_candidates_request",
+        project_id=project_id,
+        limit=limit,
+    )
+    return await query_km_stale_candidates(project_id=project_id, limit=limit)
+
+
+# =============================================================================
+# GET /api/v1/ai/governance/km-stale-owner-reviews
+# =============================================================================
+
+@router.get(
+    "/ai/governance/km-stale-owner-reviews",
+    response_model=KnowledgeStaleOwnerReviewInboxResponse,
+)
+async def get_km_stale_owner_reviews(
+    project_id: Annotated[str, Query(min_length=1, max_length=64)] = "awoooi",
+    dispatch_status: Annotated[
+        str,
+        Query(pattern="^(all|pending|dispatched|executing|succeeded|failed|skipped|cancelled)$"),
+    ] = "pending",
+    limit: Annotated[int, Query(ge=5, le=100)] = 20,
+) -> KnowledgeStaleOwnerReviewInboxResponse:
+    """
+    查詢 stale KM owner-review 工作台。
+
+    這是 read-only inbox：把 dispatch trail 與 KM priority context 合併，
+    讓 operator 可以依 P0/P1、score、batch 來源與流程階段逐筆 completion。
+    """
+    logger.debug(
+        "km_stale_owner_reviews_request",
+        project_id=project_id,
+        dispatch_status=dispatch_status,
+        limit=limit,
+    )
+    try:
+        return await query_km_stale_owner_review_inbox(
+            project_id=project_id,
+            dispatch_status=dispatch_status,
+            limit=limit,
+        )
+    except KmStaleOwnerReviewError as exc:
+        raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
+
+
+# =============================================================================
+# GET /api/v1/ai/governance/km-stale-owner-review-burndown
+# =============================================================================
+
+@router.get(
+    "/ai/governance/km-stale-owner-review-burndown",
+    response_model=KnowledgeStaleOwnerReviewBurnDownResponse,
+)
+async def get_km_stale_owner_review_burndown(
+    project_id: Annotated[str, Query(min_length=1, max_length=64)] = "awoooi",
+    limit: Annotated[int, Query(ge=1, le=100)] = 20,
+) -> KnowledgeStaleOwnerReviewBurnDownResponse:
+    """
+    查詢 stale KM owner-review 完成與 stale ratio burn-down 狀態。
+
+    這是 read-only dashboard：把 pending review、completion audit、recheck
+    snapshot 與距離治理門檻的剩餘筆數放在同一個前端面板。
+    """
+    logger.debug(
+        "km_stale_owner_review_burndown_request",
+        project_id=project_id,
+        limit=limit,
+    )
+    return await query_km_stale_owner_review_burndown(
+        project_id=project_id,
+        limit=limit,
+    )
+
+
+# =============================================================================
+# GET /api/v1/ai/governance/km-stale-owner-review-completion-queue
+# =============================================================================
+
+@router.get(
+    "/ai/governance/km-stale-owner-review-completion-queue",
+    response_model=KnowledgeStaleOwnerReviewCompletionQueueResponse,
+)
+async def get_km_stale_owner_review_completion_queue(
+    project_id: Annotated[str, Query(min_length=1, max_length=64)] = "awoooi",
+    status_bucket: Annotated[
+        str,
+        Query(pattern="^(all|ready|blocked|completed|failed|pending)$"),
+    ] = "all",
+    priority_tier: Annotated[list[str] | None, Query(alias="priority_tier")] = None,
+    recommended_completion_outcome: Annotated[
+        str,
+        Query(pattern="^(all|refresh_with_evidence|archive|supersede)$"),
+    ] = "all",
+    batch_governance_event_id: Annotated[str | None, Query(max_length=120)] = None,
+    can_preview: bool | None = None,
+    limit: Annotated[int, Query(ge=1, le=100)] = 20,
+) -> KnowledgeStaleOwnerReviewCompletionQueueResponse:
+    """
+    查詢 stale KM owner-review completion 分流。
+
+    這是 read-only queue：把 active / completed / failed dispatch 拆成
+    ready、blocked、completed、failed，讓前端呈現下一步卡點；打開頁面不寫 KM。
+    """
+    logger.debug(
+        "km_stale_owner_review_completion_queue_request",
+        project_id=project_id,
+        status_bucket=status_bucket,
+        priority_tiers=priority_tier,
+        recommended_completion_outcome=recommended_completion_outcome,
+        batch_governance_event_id=batch_governance_event_id,
+        can_preview=can_preview,
+        limit=limit,
+    )
+    try:
+        return await query_km_stale_owner_review_completion_queue(
+            project_id=project_id,
+            status_bucket=status_bucket,
+            priority_tiers=priority_tier,
+            recommended_completion_outcome=recommended_completion_outcome,
+            batch_governance_event_id=batch_governance_event_id,
+            can_preview=can_preview,
+            limit=limit,
+        )
+    except KmStaleOwnerReviewError as exc:
+        raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
+
+
+# =============================================================================
+# POST /api/v1/ai/governance/km-stale-owner-review-completion-queue/batch-preview
+# =============================================================================
+
+@router.post(
+    "/ai/governance/km-stale-owner-review-completion-queue/batch-preview",
+    response_model=KnowledgeStaleOwnerReviewCompletionBatchPreviewResponse,
+)
+async def post_km_stale_owner_review_completion_batch_preview(
+    request: KnowledgeStaleOwnerReviewCompletionBatchPreviewRequest,
+) -> KnowledgeStaleOwnerReviewCompletionBatchPreviewResponse:
+    """
+    Preview a bounded set of owner-review completion candidates.
+
+    This endpoint is intentionally dry-run only: it does not write KM, does not
+    enqueue a batch executor, and does not create governance audit rows. Each
+    item must still be completed through the single-item dry-run + owner confirm
+    endpoint.
+    """
+    logger.info(
+        "km_stale_owner_review_completion_batch_preview_request",
+        project_id=request.project_id,
+        status_bucket=request.status_bucket,
+        priority_tiers=request.priority_tiers,
+        recommended_completion_outcome=request.recommended_completion_outcome,
+        batch_governance_event_id=request.batch_governance_event_id,
+        limit=request.limit,
+        owner=request.owner,
+    )
+    try:
+        return await preview_km_stale_owner_review_completion_batch(request=request)
+    except KmStaleOwnerReviewError as exc:
+        raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
+
+
+# =============================================================================
+# POST /api/v1/ai/governance/km-stale-candidates/batch-queue-review
+# =============================================================================
+
+@router.post(
+    "/ai/governance/km-stale-candidates/batch-queue-review",
+    response_model=KnowledgeStaleOwnerReviewBatchQueueResponse,
+)
+async def post_km_stale_candidate_batch_queue_review(
+    request: KnowledgeStaleOwnerReviewBatchQueueRequest,
+) -> KnowledgeStaleOwnerReviewBatchQueueResponse:
+    """
+    將 P0/P1 stale KM 批次排入 owner review。
+
+    這個 endpoint 只建立 batch audit 與逐筆 owner-review dispatch，不改寫 KM。
+    真正 refresh / archive / supersede 仍需單筆 dry-run fingerprint + owner approval。
+    """
+    logger.info(
+        "km_stale_candidate_batch_queue_review_request",
+        project_id=request.project_id,
+        priority_tiers=request.priority_tiers,
+        limit=request.limit,
+        owner=request.owner,
+        dry_run=request.dry_run,
+    )
+    try:
+        return await batch_queue_km_stale_owner_reviews(request=request)
+    except KmStaleOwnerReviewError as exc:
+        raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
+
+
+# =============================================================================
+# POST /api/v1/ai/governance/km-stale-candidates/{entry_id}/queue-review
+# =============================================================================
+
+@router.post(
+    "/ai/governance/km-stale-candidates/{entry_id}/queue-review",
+    response_model=KnowledgeStaleOwnerReviewResponse,
+)
+async def post_km_stale_candidate_queue_review(
+    entry_id: str,
+    request: KnowledgeStaleOwnerReviewRequest,
+) -> KnowledgeStaleOwnerReviewResponse:
+    """
+    將單筆 stale KM candidate 排入 owner review。
+
+    這個 endpoint 只建立治理事件與 dispatch work item，不修改 KM 內容。
+    實際 refresh / archive / supersede 仍需 owner 在後續流程確認。
+    """
+    logger.info(
+        "km_stale_candidate_queue_review_request",
+        entry_id=entry_id,
+        owner=request.owner,
+        dry_run=request.dry_run,
+    )
+    try:
+        return await queue_km_stale_owner_review(entry_id=entry_id, request=request)
+    except KmStaleOwnerReviewError as exc:
+        raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
+
+
+# =============================================================================
+# POST /api/v1/ai/governance/km-stale-candidates/{entry_id}/complete-review
+# =============================================================================
+
+@router.post(
+    "/ai/governance/km-stale-candidates/{entry_id}/complete-review",
+    response_model=KnowledgeStaleOwnerReviewCompleteResponse,
+)
+async def post_km_stale_candidate_complete_review(
+    entry_id: str,
+    request: KnowledgeStaleOwnerReviewCompleteRequest,
+) -> KnowledgeStaleOwnerReviewCompleteResponse:
+    """
+    Owner 審核後完成 stale KM 的 refresh / archive / supersede 流程。
+
+    必須先 dry-run 取得 fingerprint；真正寫入時需 owner_approved=true。
+    後端會寫 KM、terminal audit dispatch 與 stale ratio recheck dispatch。
+    """
+    logger.info(
+        "km_stale_candidate_complete_review_request",
+        entry_id=entry_id,
+        dispatch_id=request.dispatch_id,
+        owner=request.owner,
+        review_outcome=request.review_outcome,
+        dry_run=request.dry_run,
+        owner_approved=request.owner_approved,
+    )
+    try:
+        return await complete_km_stale_owner_review(
+            entry_id=entry_id,
+            request=request,
+        )
+    except KmStaleOwnerReviewError as exc:
+        raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
+
+
 # =============================================================================
 # GET /api/v1/ai/governance/summary
 # =============================================================================

apps/api/src/api/v1/ai_slo.py

@@ -18,8 +18,15 @@ Endpoints:
 from __future__ import annotations
 
 import structlog
-from fastapi import APIRouter, Query
+from fastapi import APIRouter, HTTPException, Query
+from pydantic import BaseModel, Field
 
+from src.services.adr100_remediation_service import (
+    RemediationMode,
+    RemediationNotFoundError,
+    get_adr100_remediation_service,
+)
+from src.services.adr100_slo_status_service import get_adr100_slo_status_service
 from src.services.ai_slo_calculator import AiSloCalculator
 
 logger = structlog.get_logger(__name__)
@@ -27,6 +34,20 @@ logger = structlog.get_logger(__name__)
 router = APIRouter()
 
 
+class RemediationPreviewRequest(BaseModel):
+    """ADR-100 remediation preview request."""
+
+    work_item_id: str = Field(min_length=1)
+    mode: RemediationMode = "auto"
+
+
+class RemediationDryRunRequest(BaseModel):
+    """ADR-100 remediation dry-run request."""
+
+    work_item_id: str = Field(min_length=1)
+    mode: RemediationMode = "auto"
+
+
 @router.get("/ai/slo")
 async def get_ai_slo(
     force_refresh: bool = Query(False, description="忽略快取，強制重算"),
@@ -50,9 +71,65 @@ async def get_ai_slo(
         if cached:
             data = cached.to_dict()
             data["cache_hit"] = True
+            data["adr100"] = await get_adr100_slo_status_service().fetch_report()
             return data
 
     report = await calc.run()
     data = report.to_dict()
     data["cache_hit"] = False
+    data["adr100"] = await get_adr100_slo_status_service().fetch_report()
     return data
+
+
+@router.get("/ai/slo/remediation/preview")
+async def preview_ai_slo_remediation(
+    work_item_id: str = Query(..., min_length=1),
+    mode: RemediationMode = Query("auto"),
+) -> dict:
+    """Preview the safe remediation plan for one ADR-100 queue item."""
+
+    try:
+        return await get_adr100_remediation_service().preview(work_item_id, mode)
+    except RemediationNotFoundError as exc:
+        raise HTTPException(status_code=404, detail="remediation_work_item_not_found") from exc
+
+
+@router.post("/ai/slo/remediation/preview")
+async def preview_ai_slo_remediation_post(request: RemediationPreviewRequest) -> dict:
+    """POST variant for clients that prefer JSON bodies."""
+
+    try:
+        return await get_adr100_remediation_service().preview(
+            request.work_item_id,
+            request.mode,
+        )
+    except RemediationNotFoundError as exc:
+        raise HTTPException(status_code=404, detail="remediation_work_item_not_found") from exc
+
+
+@router.post("/ai/slo/remediation/dry-run")
+async def dry_run_ai_slo_remediation(request: RemediationDryRunRequest) -> dict:
+    """Run a read-only ADR-100 remediation dry-run."""
+
+    try:
+        return await get_adr100_remediation_service().dry_run(
+            request.work_item_id,
+            request.mode,
+        )
+    except RemediationNotFoundError as exc:
+        raise HTTPException(status_code=404, detail="remediation_work_item_not_found") from exc
+
+
+@router.get("/ai/slo/remediation/history")
+async def list_ai_slo_remediation_history(
+    limit: int = Query(50, ge=1, le=200),
+    incident_id: str | None = Query(default=None, min_length=1),
+    work_item_id: str | None = Query(default=None, min_length=1),
+) -> dict:
+    """List durable ADR-100 remediation dry-run history from alert_operation_log."""
+
+    return await get_adr100_remediation_service().history(
+        limit=limit,
+        incident_id=incident_id,
+        work_item_id=work_item_id,
+    )

apps/api/src/api/v1/dashboard.py

@@ -20,6 +20,7 @@ from pydantic import BaseModel
 from src.core.config import settings
 from src.core.logging import get_logger
 from src.core.sse import EventPublisher, EventType, SSEEvent, get_publisher
+from src.services.dashboard_metrics_service import fetch_pending_approval_count
 from src.services.host_aggregator import AggregatedStatus, HostAggregator
 
 router = APIRouter()
@@ -141,12 +142,14 @@ async def dashboard_update_loop(publisher: EventPublisher) -> None:
         try:
             # Fetch aggregated status
             status = await HostAggregator.fetch_all()
+            pending_approvals = await fetch_pending_approval_count()
 
             # Publish to all connected clients
             event = SSEEvent(
                 type=EventType.HOST_UPDATE,
                 data={
                     "overall_status": status.overall_status,
+                    "pending_approvals": pending_approvals,
                     "hosts": [
                         {
                             "ip": h.ip,
@@ -206,7 +209,9 @@ async def get_dashboard() -> DashboardResponse:
     logger.info("dashboard_fetch")
 
     status = await HostAggregator.fetch_all()
-    return aggregated_to_response(status)
+    response = aggregated_to_response(status)
+    response.pending_approvals = await fetch_pending_approval_count()
+    return response
 
 
 @router.get("/dashboard/stream")

apps/api/src/api/v1/drift.py

@@ -13,10 +13,12 @@ leWOOOgo 積木化原則:
 建立者: Claude Code (Phase 25 P2)
 """
 
+from typing import Literal
+
 from fastapi import APIRouter, BackgroundTasks, HTTPException
+from pydantic import BaseModel, Field
 
 from src.core.csrf import CSRFToken  # Phase 20: CSRF Protection
-
 from src.models.drift import (
     DriftListResponse,
     DriftReport,
@@ -28,6 +30,10 @@ from src.repositories.drift_repository import get_drift_repository
 from src.services.drift_adopt_service import get_drift_adopt_service
 from src.services.drift_analyzer import get_drift_analyzer
 from src.services.drift_detector import get_drift_detector
+from src.services.drift_fingerprint_state_service import (
+    DriftFingerprintStateNotFoundError,
+    get_drift_fingerprint_state_service,
+)
 from src.services.drift_interpreter import get_drift_interpreter
 from src.services.drift_remediator import get_drift_remediator
 from src.utils.timezone import now_taipei
@@ -37,6 +43,42 @@ router = APIRouter(prefix="/drift", tags=["drift"])
 # 2026-04-09 Claude Sonnet 4.6: B4 drift_reports 持久化 — 改用 DB repository
 
 
+class DriftFingerprintHandoffRequest(BaseModel):
+    """Record-only handoff request for a stable drift fingerprint."""
+
+    report_id: str | None = Field(default=None, min_length=1)
+    namespace: str | None = Field(default="awoooi-prod", min_length=1)
+    handoff_kind: Literal[
+        "open_pr_review",
+        "manual_investigation",
+        "zero_diff_pr_cleanup",
+    ] = "open_pr_review"
+    pr_url: str | None = Field(default=None, min_length=1)
+    note: str | None = Field(default=None, max_length=500)
+
+
+class DriftFingerprintRemediationRequest(BaseModel):
+    """Record-only remediation request for a stable drift fingerprint."""
+
+    report_id: str | None = Field(default=None, min_length=1)
+    namespace: str | None = Field(default="awoooi-prod", min_length=1)
+    remediation_kind: Literal[
+        "live_env_rollback",
+        "git_adopted",
+        "git_rollback",
+        "zero_diff_pr_cleanup",
+        "manual_noop",
+    ] = "live_env_rollback"
+    remediation_status: Literal[
+        "executed_unverified",
+        "verified_no_drift",
+        "verification_failed",
+    ] | None = None
+    verification_report_id: str | None = Field(default=None, min_length=1)
+    note: str | None = Field(default=None, max_length=1000)
+    commands_summary: list[str] = Field(default_factory=list, max_length=12)
+
+
 @router.post("/scan", response_model=DriftScanResponse, summary="觸發漂移掃描")
 async def trigger_drift_scan(
     request: DriftScanRequest,
@@ -99,6 +141,72 @@ async def list_drift_reports() -> DriftListResponse:
     return DriftListResponse(items=items, total=len(items))
 
 
+@router.get("/fingerprints/state", summary="查詢 Config Drift fingerprint 狀態")
+async def get_drift_fingerprint_state(
+    report_id: str | None = None,
+    namespace: str | None = "awoooi-prod",
+) -> dict:
+    """
+    以 stable fingerprint 聚合漂移狀態。
+
+    此 endpoint 只建立 read model：重複次數、PR 狀態、是否零 diff、
+    人工交接歷史與下一步。它不修改 drift / incident / auto-repair 狀態。
+    """
+    svc = get_drift_fingerprint_state_service()
+    try:
+        return await svc.get_state(report_id=report_id, namespace=namespace)
+    except DriftFingerprintStateNotFoundError as exc:
+        raise HTTPException(status_code=404, detail="drift_report_not_found") from exc
+
+
+@router.post("/fingerprints/handoff", summary="記錄 Config Drift fingerprint 交接")
+async def record_drift_fingerprint_handoff(
+    request: DriftFingerprintHandoffRequest,
+) -> dict:
+    """
+    記錄 stable fingerprint 已轉人工 / PR review 的歷史證據。
+
+    安全邊界：只寫 alert_operation_log / timeline_events，不修改 drift 狀態、
+    incident 狀態、自動修復結果，不建立外部 ticket，也不 merge PR。
+    """
+    svc = get_drift_fingerprint_state_service()
+    try:
+        return await svc.record_handoff(
+            report_id=request.report_id,
+            namespace=request.namespace,
+            handoff_kind=request.handoff_kind,
+            pr_url=request.pr_url,
+            note=request.note,
+        )
+    except DriftFingerprintStateNotFoundError as exc:
+        raise HTTPException(status_code=404, detail="drift_report_not_found") from exc
+
+
+@router.post("/fingerprints/remediation", summary="記錄 Config Drift fingerprint 修復")
+async def record_drift_fingerprint_remediation(
+    request: DriftFingerprintRemediationRequest,
+) -> dict:
+    """
+    記錄 stable fingerprint 已完成的修復 / 驗證證據。
+
+    安全邊界：只寫 alert_operation_log / timeline_events，不修改 drift 狀態、
+    incident 狀態、自動修復結果，不建立外部 ticket，也不執行 kubectl。
+    """
+    svc = get_drift_fingerprint_state_service()
+    try:
+        return await svc.record_remediation(
+            report_id=request.report_id,
+            namespace=request.namespace,
+            remediation_kind=request.remediation_kind,
+            remediation_status=request.remediation_status,
+            verification_report_id=request.verification_report_id,
+            note=request.note,
+            commands_summary=request.commands_summary,
+        )
+    except DriftFingerprintStateNotFoundError as exc:
+        raise HTTPException(status_code=404, detail="drift_report_not_found") from exc
+
+
 @router.post("/reports/{report_id}/rollback", summary="覆蓋回 Git 狀態")
 async def rollback_drift(report_id: str, _csrf_token: CSRFToken) -> dict:  # Phase 20: CSRF Protection (驗證用，不需要使用值)
     """

apps/api/src/api/v1/gitea_webhook.py

@@ -418,7 +418,9 @@ async def _send_gitea_notification(
             logger.debug("gitea_tg_skipped", reason="Bot token not configured")
             return
 
-        from src.services.telegram_gateway import get_telegram_gateway  # type: ignore[import]
+        from src.services.telegram_gateway import (
+            get_telegram_gateway,  # type: ignore[import]
+        )
         gateway = get_telegram_gateway()
         await gateway.initialize()
         await gateway.send_alert_notification(message)
@@ -502,15 +504,22 @@ async def handle_pull_request(
     review_id = f"gitea-pr-{payload.repository.id}-{pr.number}-{uuid.uuid4().hex[:8]}"
 
     # 背景執行審查 (委派給 Service)
-    service = get_gitea_webhook_service()
-    background_tasks.add_task(
-        service.review_pull_request,
-        repo=payload.repository,
-        pr=pr,
-        sender=payload.sender,
-        review_id=review_id,
-        action=payload.action,
-    )
+    if settings.MOCK_MODE:
+        logger.info(
+            "gitea_pr_review_background_skipped_mock_mode",
+            review_id=review_id,
+            repo=payload.repository.full_name,
+        )
+    else:
+        service = get_gitea_webhook_service()
+        background_tasks.add_task(
+            service.review_pull_request,
+            repo=payload.repository,
+            pr=pr,
+            sender=payload.sender,
+            review_id=review_id,
+            action=payload.action,
+        )
 
     logger.info(
         "gitea_pr_review_scheduled",
@@ -561,17 +570,24 @@ async def handle_push(
     review_id = f"gitea-push-{payload.repository.id}-{payload.after[:8]}-{uuid.uuid4().hex[:8]}"
 
     # 背景執行審查 (委派給 Service)
-    service = get_gitea_webhook_service()
-    background_tasks.add_task(
-        service.review_push,
-        repo=payload.repository,
-        commits=commits,
-        sender=payload.sender,
-        review_id=review_id,
-        ref=ref,
-        before_sha=payload.before,
-        after_sha=payload.after,
-    )
+    if settings.MOCK_MODE:
+        logger.info(
+            "gitea_push_review_background_skipped_mock_mode",
+            review_id=review_id,
+            repo=payload.repository.full_name,
+        )
+    else:
+        service = get_gitea_webhook_service()
+        background_tasks.add_task(
+            service.review_push,
+            repo=payload.repository,
+            commits=commits,
+            sender=payload.sender,
+            review_id=review_id,
+            ref=ref,
+            before_sha=payload.before,
+            after_sha=payload.after,
+        )
 
     logger.info(
         "gitea_push_review_scheduled",

apps/api/src/api/v1/health.py

@@ -11,7 +11,7 @@ Endpoints:
 Components Checked:
 - PostgreSQL (192.168.0.188:5432)
 - Redis (192.168.0.188:6380)
-- Ollama (settings.OLLAMA_URL / ADR-110 provider pool)
+- Ollama ADR-110 provider pool (GCP-A -> GCP-B -> 111)
 - OpenClaw (192.168.0.188:8089)
 - SigNoz (192.168.0.188:3301)
 """
@@ -26,9 +26,16 @@ from pydantic import BaseModel
 from src.core.config import settings
 from src.core.logging import get_logger
 from src.services.health_check_service import get_health_check_service
+from src.services.ollama_endpoint_circuit_breaker import (
+    get_ollama_endpoint_cooldown_remaining_seconds,
+    record_ollama_endpoint_failure,
+    record_ollama_endpoint_success,
+)
+from src.services.ollama_endpoint_resolver import resolve_ollama_order
 
 router = APIRouter()
 logger = get_logger("awoooi.health")
+CORE_COMPONENTS = ("api", "postgresql", "redis", "ollama", "openclaw", "signoz")
 
 
 # =============================================================================
@@ -40,6 +47,11 @@ class ComponentHealth(BaseModel):
     status: Literal["up", "down", "degraded"]
     latency_ms: float | None = None
     error: str | None = None
+    provider_name: str | None = None
+    diagnosis_code: str | None = None
+    retry_after_seconds: float | None = None
+    cooldown_remaining_seconds: float | None = None
+    is_cooldown: bool = False
 
 
 class HealthResponse(BaseModel):
@@ -50,6 +62,7 @@ class HealthResponse(BaseModel):
     mock_mode: bool
     timestamp: datetime
     components: dict[str, ComponentHealth]
+    ollama_route_order: list[str] = []
 
 
 # =============================================================================
@@ -106,8 +119,125 @@ async def check_redis() -> ComponentHealth:
 
 
 async def check_ollama() -> ComponentHealth:
-    """Async Ollama health check via /api/tags"""
-    return await _http_health_check("ollama", settings.OLLAMA_URL, "/api/tags")
+    """Async aggregate Ollama health check via ADR-110 provider chain."""
+    aggregate, _details = await check_ollama_provider_chain()
+    return aggregate
+
+
+async def check_ollama_provider_chain() -> tuple[ComponentHealth, dict[str, ComponentHealth]]:
+    """
+    Check the full Ollama provider chain.
+
+    The aggregate ``ollama`` component represents route availability:
+    - up: GCP-A is reachable
+    - degraded: GCP-A is unavailable but GCP-B or 111 is reachable
+    - down: no configured Ollama endpoint is reachable
+    """
+    selections = tuple(
+        selection
+        for selection in resolve_ollama_order("healthcheck")
+        if selection.url and selection.provider_name != "ollama_unconfigured"
+    )
+    if not selections:
+        aggregate = ComponentHealth(
+            status="down",
+            error="no Ollama endpoints configured",
+        )
+        return aggregate, {}
+
+    checked = await asyncio.gather(
+        *(
+            _ollama_endpoint_health_check(selection.provider_name, selection.url)
+            for selection in selections
+        )
+    )
+    details = {
+        selection.provider_name: result
+        for selection, result in zip(selections, checked, strict=False)
+    }
+
+    primary = selections[0]
+    primary_status = details[primary.provider_name].status
+    if primary.provider_name == "ollama_gcp_a" and primary_status == "up":
+        return details[primary.provider_name], details
+
+    first_available = next(
+        (
+            selection
+            for selection in selections
+            if details[selection.provider_name].status == "up"
+        ),
+        None,
+    )
+    if first_available:
+        fallback = details[first_available.provider_name]
+        return (
+            ComponentHealth(
+                status="degraded",
+                latency_ms=fallback.latency_ms,
+                error=f"primary unavailable; fallback active: {first_available.provider_name}",
+            ),
+            details,
+        )
+
+    errors = ", ".join(
+        f"{provider}={health.error or health.status}"
+        for provider, health in details.items()
+    )
+    return (
+        ComponentHealth(
+            status="down",
+            error=f"all Ollama endpoints unavailable: {errors}",
+        ),
+        details,
+    )
+
+
+async def _ollama_endpoint_health_check(name: str, url: str) -> ComponentHealth:
+    cooldown_remaining = get_ollama_endpoint_cooldown_remaining_seconds(url)
+    if cooldown_remaining > 0:
+        return ComponentHealth(
+            status="down",
+            error=f"recent endpoint failure cooldown: {cooldown_remaining:.0f}s",
+            provider_name=name,
+            diagnosis_code="endpoint_cooldown",
+            retry_after_seconds=round(cooldown_remaining, 1),
+            cooldown_remaining_seconds=round(cooldown_remaining, 1),
+            is_cooldown=True,
+        )
+
+    result = await _http_health_check(name, url, "/api/tags")
+    result.provider_name = name
+    if result.status == "up":
+        result.diagnosis_code = "endpoint_reachable"
+        record_ollama_endpoint_success(url)
+    else:
+        result.diagnosis_code = _classify_ollama_endpoint_failure(name, result.error)
+        record_ollama_endpoint_failure(url)
+    return result
+
+
+def _classify_ollama_endpoint_failure(
+    provider_name: str,
+    error: str | None,
+) -> str:
+    """Return a stable diagnosis code for UI/alert rendering."""
+    normalized_error = (error or "").lower()
+    if "cooldown" in normalized_error:
+        return "endpoint_cooldown"
+    if "502" in normalized_error or "bad gateway" in normalized_error:
+        return (
+            "local_proxy_upstream_unreachable"
+            if provider_name == "ollama_local"
+            else "proxy_upstream_unreachable"
+        )
+    if "timeout" in normalized_error:
+        return "endpoint_timeout"
+    if "connection refused" in normalized_error:
+        return "endpoint_connection_refused"
+    if "no route to host" in normalized_error or "network is unreachable" in normalized_error:
+        return "endpoint_network_unreachable"
+    return "endpoint_unreachable"
 
 
 async def check_openclaw() -> ComponentHealth:
@@ -120,6 +250,30 @@ async def check_signoz() -> ComponentHealth:
     return await _http_health_check("signoz", settings.SIGNOZ_URL, "/api/v1/health")
 
 
+def _determine_overall_status(
+    components: dict[str, ComponentHealth],
+) -> Literal["healthy", "degraded", "unhealthy"]:
+    """Determine overall health from core aggregate components only."""
+    statuses = [
+        components[name].status
+        for name in CORE_COMPONENTS
+        if name in components
+    ]
+    down_count = statuses.count("down")
+    degraded_count = statuses.count("degraded")
+
+    critical_down = (
+        components.get("postgresql", ComponentHealth(status="down")).status == "down"
+        or components.get("redis", ComponentHealth(status="down")).status == "down"
+    )
+
+    if critical_down or down_count >= 3:
+        return "unhealthy"
+    if down_count >= 1 or degraded_count > 0:
+        return "degraded"
+    return "healthy"
+
+
 # =============================================================================
 # Endpoints
 # =============================================================================
@@ -142,34 +296,28 @@ async def get_health() -> HealthResponse:
     results = await asyncio.gather(
         check_postgresql(),
         check_redis(),
-        check_ollama(),
+        check_ollama_provider_chain(),
         check_openclaw(),
         check_signoz(),
     )
 
+    ollama_aggregate, ollama_details = results[2]
     components = {
         "api": ComponentHealth(status="up", latency_ms=0.0),
         "postgresql": results[0],
         "redis": results[1],
-        "ollama": results[2],
+        "ollama": ollama_aggregate,
         "openclaw": results[3],
         "signoz": results[4],
     }
+    components.update(ollama_details)
 
-    # Determine overall status
-    statuses = [c.status for c in components.values()]
-    down_count = statuses.count("down")
-    degraded_count = statuses.count("degraded")
-
-    # Critical services: postgresql, redis
-    critical_down = components["postgresql"].status == "down" or components["redis"].status == "down"
-
-    if critical_down or down_count >= 3:
-        overall_status: Literal["healthy", "degraded", "unhealthy"] = "unhealthy"
-    elif down_count >= 1 or degraded_count > 0:
-        overall_status = "degraded"
-    else:
-        overall_status = "healthy"
+    overall_status = _determine_overall_status(components)
+    ollama_route_order = [
+        selection.provider_name
+        for selection in resolve_ollama_order("healthcheck")
+        if selection.url and selection.provider_name != "ollama_unconfigured"
+    ]
 
     logger.info(
         "health_check_complete",
@@ -185,6 +333,7 @@ async def get_health() -> HealthResponse:
         mock_mode=settings.MOCK_MODE,
         timestamp=datetime.now(UTC),
         components=components,
+        ollama_route_order=ollama_route_order,
     )
 
 

apps/api/src/api/v1/incidents.py

@@ -134,6 +134,7 @@ class IncidentTimelineResponse(BaseModel):
     timeline: list[IncidentTimelineStage] = Field(default_factory=list)
     events: list[IncidentTimelineEvent] = Field(default_factory=list)
     ascii_timeline: str
+    reconciliation: dict[str, Any] = Field(default_factory=dict)
 
 
 # =============================================================================

apps/api/src/api/v1/platform/__init__.py

@@ -13,9 +13,11 @@ from src.api.v1.platform.events import router as events_router
 from src.api.v1.platform.operator_runs import router as operator_runs_router
 from src.api.v1.platform.runs import router as runs_router
 from src.api.v1.platform.tenants import router as tenants_router
+from src.api.v1.platform.truth_chain import router as truth_chain_router
 
 router = APIRouter()
 router.include_router(events_router)
+router.include_router(truth_chain_router)
 # 2026-05-06 Codex: FastAPI 依註冊順序比對路由。Operator Console 的
 # `/runs/list` 必須排在 `/runs/{run_id}` 前面，否則 `list` 會被當成
 # run_id，造成前端 Run 監控頁 HTTP 422。

apps/api/src/api/v1/platform/events.py

@@ -6,13 +6,32 @@ AwoooP Operator Console — Channel Events API
 
 from __future__ import annotations
 
-from datetime import datetime
-from typing import Any
+from datetime import UTC, datetime
+from typing import Annotated, Any, Literal
 from uuid import UUID
 
-from fastapi import APIRouter, Query
-from pydantic import BaseModel
+from fastapi import APIRouter, Depends, HTTPException, Query
+from pydantic import BaseModel, Field
 
+from src.core.awooop_operator_auth import (
+    AwoooPOperatorPrincipal,
+    verify_awooop_operator,
+)
+from src.services.channel_event_dossier_service import (
+    RecurrenceWorkItemHandoffKind,
+    RecurrenceWorkItemMode,
+    RecurrenceWorkItemNotFoundError,
+    SourceCorrelationReviewDecision,
+    fetch_channel_event_dossier,
+    fetch_channel_event_dossier_coverage,
+    fetch_channel_event_dossier_recurrence,
+    fetch_recurrence_work_item_dry_run,
+    fetch_recurrence_work_item_handoff,
+    fetch_recurrence_work_item_preview,
+    fetch_source_correlation_apply,
+    fetch_source_correlation_review_decision,
+)
+from src.services.channel_hub import record_external_alert_event
 from src.services.platform_operator_service import list_recent_channel_events
 
 router = APIRouter()
@@ -35,6 +54,513 @@ class RecentEventsResponse(BaseModel):
     limit: int
 
 
+class ChannelEventDossierItem(BaseModel):
+    event_id: UUID
+    project_id: str
+    channel_type: str
+    provider: str | None
+    stage: str
+    provider_event_id: str
+    content_preview: str | None
+    content_redacted: str | None
+    has_redacted_content: bool
+    redaction_version: str | None
+    source_url: str | None
+    content_sha256: str | None
+    content_length: int | None
+    source_refs: dict[str, Any]
+    source_ref_count: int
+    log_correlation: dict[str, Any]
+    alertname: str | None
+    severity: str | None
+    namespace: str | None
+    target_resource: str | None
+    fingerprint: str | None
+    is_duplicate: bool
+    provider_ts: datetime | None
+    received_at: datetime
+
+
+class ChannelEventDossierSummary(BaseModel):
+    source_count: int
+    duplicate_total: int
+    redacted_total: int
+    source_ref_total: int
+
+
+class ChannelEventDossierResponse(BaseModel):
+    events: list[ChannelEventDossierItem]
+    total: int
+    limit: int
+    summary: ChannelEventDossierSummary
+
+
+class ChannelEventProviderCoverage(BaseModel):
+    provider: str
+    total: int
+    duplicate_total: int
+    redacted_total: int
+    source_ref_total: int
+    missing_source_refs_total: int
+    sentry_ref_total: int
+    signoz_ref_total: int
+    alert_ref_total: int
+    latest_received_at: datetime | None
+
+
+class ChannelEventDossierCoverageSummary(BaseModel):
+    source_count: int
+    source_envelope_total: int
+    missing_source_envelope_total: int
+    with_source_refs_total: int
+    missing_source_refs_total: int
+    duplicate_total: int
+    redacted_total: int
+    source_ref_total: int
+    sentry_ref_total: int
+    signoz_ref_total: int
+    alert_ref_total: int
+    latest_received_at: datetime | None
+
+
+class ChannelEventDossierCoverageResponse(BaseModel):
+    project_id: str
+    limit: int
+    summary: ChannelEventDossierCoverageSummary
+    providers: list[ChannelEventProviderCoverage]
+
+
+SourceProviderName = Literal["sentry", "signoz"]
+
+
+class SourceProviderHeartbeatRequest(BaseModel):
+    """Low-noise freshness heartbeat for external source-provider mirrors."""
+
+    project_id: str = Field(default="awoooi", min_length=1, max_length=64)
+    providers: list[SourceProviderName] = Field(
+        default_factory=lambda: ["sentry", "signoz"],
+        min_length=1,
+        max_length=2,
+    )
+    reason: str = Field(
+        default="scheduled_provider_freshness_smoke",
+        min_length=1,
+        max_length=120,
+    )
+    run_ref: str | None = Field(default=None, max_length=120)
+
+
+class SourceProviderHeartbeatItem(BaseModel):
+    provider: SourceProviderName
+    event_id: str
+    conversation_event_id: UUID
+
+
+class SourceProviderHeartbeatResponse(BaseModel):
+    status: str
+    project_id: str
+    items: list[SourceProviderHeartbeatItem]
+
+
+class ChannelEventRecurrenceSummary(BaseModel):
+    source_event_total: int
+    recurrence_group_total: int
+    recurrent_group_total: int
+    duplicate_event_total: int
+    linked_run_total: int
+    unlinked_event_total: int
+    auto_repair_linked_total: int = 0
+    verified_repair_group_total: int = 0
+    open_work_item_group_total: int = 0
+    manual_gate_group_total: int = 0
+    automation_gap_group_total: int = 0
+    failed_repair_group_total: int = 0
+    source_correlation_review_group_total: int = 0
+    source_correlation_decision_recorded_group_total: int = 0
+    source_correlation_applied_group_total: int = 0
+    latest_received_at: datetime | None
+
+
+class ChannelEventRecurrenceItem(BaseModel):
+    recurrence_key: str
+    provider: str | None
+    alertname: str | None
+    severity: str | None
+    namespace: str | None
+    target_resource: str | None
+    fingerprint: str | None
+    latest_stage: str | None = None
+    latest_event_id: UUID | None
+    latest_provider_event_id: str | None
+    latest_content_preview: str | None
+    latest_run_id: UUID | None
+    latest_run_state: str | None
+    latest_agent_id: str | None
+    latest_incident_id: str | None = None
+    incident_ids: list[str] = Field(default_factory=list)
+    repair_summary: dict[str, Any] | None = None
+    work_item: dict[str, Any] | None = None
+    source_correlation_review: dict[str, Any] | None = None
+    source_correlation_apply: dict[str, Any] | None = None
+    occurrence_total: int
+    duplicate_total: int
+    linked_run_total: int
+    source_ref_total: int
+    missing_source_refs_total: int
+    sentry_ref_total: int
+    signoz_ref_total: int
+    alert_ref_total: int
+    stage_counts: dict[str, int] = Field(default_factory=dict)
+    run_state_counts: dict[str, int]
+    first_received_at: datetime | None
+    latest_received_at: datetime | None
+
+
+class ChannelEventRecurrenceResponse(BaseModel):
+    project_id: str
+    limit: int
+    summary: ChannelEventRecurrenceSummary
+    items: list[ChannelEventRecurrenceItem]
+
+
+class RecurrenceWorkItemDryRunRequest(BaseModel):
+    """AwoooP recurrence work item dry-run request."""
+
+    project_id: str | None = Field(default=None, min_length=1)
+    work_item_id: str = Field(min_length=1)
+    mode: RecurrenceWorkItemMode = "auto"
+    provider: str | None = Field(default=None, min_length=1)
+    limit: int = Field(default=300, ge=1, le=300)
+
+
+class RecurrenceWorkItemHandoffRequest(BaseModel):
+    """AwoooP recurrence work item handoff request."""
+
+    project_id: str | None = Field(default=None, min_length=1)
+    work_item_id: str = Field(min_length=1)
+    mode: RecurrenceWorkItemMode = "auto"
+    handoff_kind: RecurrenceWorkItemHandoffKind = "ticket_proposal"
+    provider: str | None = Field(default=None, min_length=1)
+    limit: int = Field(default=300, ge=1, le=300)
+
+
+class SourceCorrelationReviewDecisionRequest(BaseModel):
+    """Record-only source evidence review decision."""
+
+    project_id: str | None = Field(default=None, min_length=1)
+    work_item_id: str = Field(min_length=1)
+    decision: SourceCorrelationReviewDecision
+    target_incident_id: str | None = Field(default=None, min_length=1, max_length=30)
+    reviewer_id: str = Field(default="operator_console", min_length=1, max_length=100)
+    operator_note: str | None = Field(default=None, max_length=500)
+    provider: str | None = Field(default=None, min_length=1)
+    limit: int = Field(default=300, ge=1, le=300)
+
+
+class SourceCorrelationApplyRequest(BaseModel):
+    """Append-only source evidence link apply request."""
+
+    project_id: str | None = Field(default=None, min_length=1)
+    work_item_id: str = Field(min_length=1)
+    reviewer_id: str = Field(default="operator_console", min_length=1, max_length=100)
+    operator_note: str | None = Field(default=None, max_length=500)
+    provider: str | None = Field(default=None, min_length=1)
+    limit: int = Field(default=300, ge=1, le=300)
+
+
+@router.get(
+    "/events/dossier",
+    response_model=ChannelEventDossierResponse,
+    summary="查詢 Channel Event 來源卷宗",
+    description=(
+        "返回 redacted inbound source envelope，供 AwoooP Run Detail 顯示"
+        "告警來源、source refs、Sentry / SignOz / Alertmanager 關聯與去重狀態。"
+    ),
+)
+async def get_event_dossier(
+    project_id: str | None = Query(None, description="租戶 ID（可選）"),
+    run_id: UUID | None = Query(None, description="Run ID（可選）"),
+    provider_event_id: str | None = Query(
+        None, description="provider_event_id（可選）"
+    ),
+    limit: int = Query(20, ge=1, le=50, description="最多返回筆數"),
+) -> dict[str, Any]:
+    return await fetch_channel_event_dossier(
+        project_id=project_id,
+        run_id=run_id,
+        provider_event_id=provider_event_id,
+        limit=limit,
+    )
+
+
+@router.get(
+    "/events/dossier/coverage",
+    response_model=ChannelEventDossierCoverageResponse,
+    summary="查詢 Channel Event 來源卷宗覆蓋率",
+    description=(
+        "返回近期 inbound event 的 source_envelope / source_refs / 去重 / "
+        "Sentry / SignOz 關聯覆蓋率，供 AwoooP Run List 顯示告警是否已入庫。"
+    ),
+)
+async def get_event_dossier_coverage(
+    project_id: str | None = Query(None, description="租戶 ID（可選）"),
+    provider: str | None = Query(
+        None, description="provider（可選，如 sentry / signoz）"
+    ),
+    limit: int = Query(100, ge=1, le=200, description="最多納入統計筆數"),
+) -> dict[str, Any]:
+    return await fetch_channel_event_dossier_coverage(
+        project_id=project_id,
+        provider=provider,
+        limit=limit,
+    )
+
+
+@router.post(
+    "/events/dossier/provider-heartbeat",
+    response_model=SourceProviderHeartbeatResponse,
+    summary="寫入 Sentry / SignOz 來源卷宗 freshness heartbeat",
+    description=(
+        "受 AwoooP operator key 保護的低噪音 smoke。只寫入來源卷宗與"
+        "completed shadow run，不建立 Incident、不送 Telegram、不宣稱真實上游告警。"
+    ),
+)
+async def create_source_provider_heartbeat(
+    payload: SourceProviderHeartbeatRequest,
+    operator: Annotated[
+        AwoooPOperatorPrincipal,
+        Depends(verify_awooop_operator),
+    ],
+) -> dict[str, Any]:
+    timestamp = datetime.now(UTC).strftime("%Y%m%dT%H%M%SZ")
+    items: list[dict[str, Any]] = []
+
+    for provider in payload.providers:
+        event_id = f"heartbeat-{timestamp}"
+        event_uuid = await record_external_alert_event(
+            project_id=payload.project_id,
+            provider=provider,
+            event_id=event_id,
+            stage="heartbeat",
+            title="SourceProviderHeartbeat",
+            severity="info",
+            namespace="awoooi-prod",
+            target_resource="source-provider-ingestion",
+            fingerprint=f"source-provider-heartbeat:{provider}",
+            labels={
+                "provider": provider,
+                "synthetic": "true",
+                "alert_category": "alertchain_provider_freshness",
+                "telegram": "not_sent",
+                "incident": "not_created",
+            },
+            annotations={
+                "summary": (
+                    "Low-noise provider freshness smoke; verifies AwoooP "
+                    "source dossier ingestion without creating an incident."
+                ),
+                "reason": payload.reason,
+            },
+            payload={
+                "reason": payload.reason,
+                "run_ref": payload.run_ref,
+                "operator_id": operator.operator_id,
+                "auth_method": operator.auth_method,
+                "synthetic": True,
+                "side_effects": {
+                    "incident_created": False,
+                    "telegram_sent": False,
+                    "approval_created": False,
+                },
+            },
+        )
+        if event_uuid is None:
+            raise HTTPException(
+                status_code=500,
+                detail=f"{provider} provider heartbeat was not recorded",
+            )
+        items.append(
+            {
+                "provider": provider,
+                "event_id": event_id,
+                "conversation_event_id": event_uuid,
+            }
+        )
+
+    return {
+        "status": "recorded",
+        "project_id": payload.project_id,
+        "items": items,
+    }
+
+
+@router.get(
+    "/events/dossier/recurrence",
+    response_model=ChannelEventRecurrenceResponse,
+    summary="查詢 Channel Event 重複發生與關聯 Run 狀態",
+    description=(
+        "將近期 inbound source events 依 fingerprint / alertname / namespace / target 分組，"
+        "顯示重複發生次數、去重數、source refs 與最新 linked run 狀態。"
+    ),
+)
+async def get_event_dossier_recurrence(
+    project_id: str | None = Query(None, description="租戶 ID（可選）"),
+    provider: str | None = Query(
+        None, description="provider（可選，如 alertmanager / sentry / signoz）"
+    ),
+    limit: int = Query(100, ge=1, le=300, description="最多納入統計筆數"),
+) -> dict[str, Any]:
+    return await fetch_channel_event_dossier_recurrence(
+        project_id=project_id,
+        provider=provider,
+        limit=limit,
+    )
+
+
+@router.get(
+    "/events/dossier/recurrence/work-item/preview",
+    summary="預覽重複告警工作項的安全處理計畫",
+    description=(
+        "依 recurrence read model 找出指定 work_item，返回下一步、pre-flight checks "
+        "與 read-only / no-write 保證；不修改 incident、auto-repair 或 ticket 狀態。"
+    ),
+)
+async def preview_event_recurrence_work_item(
+    work_item_id: str = Query(..., min_length=1, description="recurrence work_item_id"),
+    project_id: str | None = Query(None, description="租戶 ID（可選）"),
+    provider: str | None = Query(
+        None, description="provider（可選，如 alertmanager / sentry / signoz）"
+    ),
+    mode: RecurrenceWorkItemMode = Query("auto", description="預覽模式"),
+    limit: int = Query(300, ge=1, le=300, description="最多納入統計筆數"),
+) -> dict[str, Any]:
+    try:
+        return await fetch_recurrence_work_item_preview(
+            project_id=project_id,
+            work_item_id=work_item_id,
+            mode=mode,
+            provider=provider,
+            limit=limit,
+        )
+    except RecurrenceWorkItemNotFoundError as exc:
+        raise HTTPException(
+            status_code=404,
+            detail="recurrence_work_item_not_found",
+        ) from exc
+
+
+@router.post(
+    "/events/dossier/recurrence/work-item/dry-run",
+    summary="乾跑重複告警工作項的安全處理流程",
+    description=(
+        "依 recurrence read model 產生 dry-run 結果並寫入 pre-flight history，"
+        "但不修改 incident、auto-repair 或 ticket 狀態。"
+    ),
+)
+async def dry_run_event_recurrence_work_item(
+    request: RecurrenceWorkItemDryRunRequest,
+) -> dict[str, Any]:
+    try:
+        return await fetch_recurrence_work_item_dry_run(
+            project_id=request.project_id,
+            work_item_id=request.work_item_id,
+            mode=request.mode,
+            provider=request.provider,
+            limit=request.limit,
+        )
+    except RecurrenceWorkItemNotFoundError as exc:
+        raise HTTPException(
+            status_code=404,
+            detail="recurrence_work_item_not_found",
+        ) from exc
+
+
+@router.post(
+    "/events/dossier/recurrence/work-item/handoff",
+    summary="記錄重複告警工作項的交接提案",
+    description=(
+        "依 recurrence read model 與 dry-run 結果記錄 ticket proposal / 人工接手歷史，"
+        "但不修改 incident、auto-repair 或外部 ticket 狀態。"
+    ),
+)
+async def handoff_event_recurrence_work_item(
+    request: RecurrenceWorkItemHandoffRequest,
+) -> dict[str, Any]:
+    try:
+        return await fetch_recurrence_work_item_handoff(
+            project_id=request.project_id,
+            work_item_id=request.work_item_id,
+            mode=request.mode,
+            handoff_kind=request.handoff_kind,
+            provider=request.provider,
+            limit=request.limit,
+        )
+    except RecurrenceWorkItemNotFoundError as exc:
+        raise HTTPException(
+            status_code=404,
+            detail="recurrence_work_item_not_found",
+        ) from exc
+
+
+@router.post(
+    "/events/dossier/recurrence/source-correlation/review",
+    summary="記錄來源證據與 Incident 配對審核結果",
+    description=(
+        "針對 source_correlation_review work item 記錄 operator 審核決定。"
+        "本 API 僅寫入 alert_operation_log / 可選 timeline_events，"
+        "不修改 Incident 狀態、不回寫 source event、不建立外部 ticket。"
+    ),
+)
+async def review_source_correlation_work_item(
+    request: SourceCorrelationReviewDecisionRequest,
+) -> dict[str, Any]:
+    try:
+        return await fetch_source_correlation_review_decision(
+            project_id=request.project_id,
+            work_item_id=request.work_item_id,
+            decision=request.decision,
+            target_incident_id=request.target_incident_id,
+            reviewer_id=request.reviewer_id,
+            operator_note=request.operator_note,
+            provider=request.provider,
+            limit=request.limit,
+        )
+    except RecurrenceWorkItemNotFoundError as exc:
+        raise HTTPException(
+            status_code=404,
+            detail="recurrence_work_item_not_found",
+        ) from exc
+
+
+@router.post(
+    "/events/dossier/recurrence/source-correlation/apply",
+    summary="套用已確認的來源證據與 Incident 配對",
+    description=(
+        "只接受已寫入 accepted review 的 source_correlation_review work item。"
+        "成功時以 append-only 方式新增 source_correlation_linked 來源事件，"
+        "並寫入 alert_operation_log / timeline_events。"
+        "不修改 Incident 狀態、不修改 auto-repair 結果、不建立外部 ticket。"
+    ),
+)
+async def apply_source_correlation_work_item(
+    request: SourceCorrelationApplyRequest,
+) -> dict[str, Any]:
+    try:
+        return await fetch_source_correlation_apply(
+            project_id=request.project_id,
+            work_item_id=request.work_item_id,
+            reviewer_id=request.reviewer_id,
+            operator_note=request.operator_note,
+            provider=request.provider,
+            limit=request.limit,
+        )
+    except RecurrenceWorkItemNotFoundError as exc:
+        raise HTTPException(
+            status_code=404,
+            detail="recurrence_work_item_not_found",
+        ) from exc
+
+
 @router.get(
     "/events/recent",
     response_model=RecentEventsResponse,
@@ -47,7 +573,9 @@ class RecentEventsResponse(BaseModel):
 async def list_recent_events(
     project_id: str | None = Query(None, description="租戶 ID（可選）"),
     channel_type: str | None = Query(None, description="通道類型（可選）"),
-    provider_prefix: str | None = Query(None, description="provider_event_id 前綴（可選）"),
+    provider_prefix: str | None = Query(
+        None, description="provider_event_id 前綴（可選）"
+    ),
     limit: int = Query(20, ge=1, le=100, description="最多返回筆數"),
 ) -> dict[str, Any]:
     return await list_recent_channel_events(

apps/api/src/api/v1/platform/operator_runs.py

@@ -25,12 +25,24 @@ from src.core.awooop_operator_auth import (
 from src.services.platform_operator_service import (
     decide_approval as decide_approval_svc,
 )
+from src.services.platform_operator_service import (
+    get_ai_route_status as get_ai_route_status_svc,
+)
+from src.services.platform_operator_service import (
+    get_awooop_status_chain as get_awooop_status_chain_svc,
+)
 from src.services.platform_operator_service import (
     get_run_detail as get_run_detail_svc,
 )
+from src.services.platform_operator_service import (
+    list_cicd_events as list_cicd_events_svc,
+)
 from src.services.platform_operator_service import (
     list_approvals as list_approvals_svc,
 )
+from src.services.platform_operator_service import (
+    list_callback_replies as list_callback_replies_svc,
+)
 from src.services.platform_operator_service import (
     list_runs as list_runs_svc,
 )
@@ -51,6 +63,8 @@ class RunItem(BaseModel):
     step_count: int
     created_at: datetime
     timeout_at: datetime | None
+    remediation_summary: dict[str, Any] | None = None
+    callback_reply_summary: dict[str, Any] | None = None
 
 
 class ListRunsResponse(BaseModel):
@@ -60,12 +74,167 @@ class ListRunsResponse(BaseModel):
     per_page: int
 
 
+class OperatorSummaryCacheInfo(BaseModel):
+    schema_version: str = "operator_summary_cache_v1"
+    status: str
+    source: str
+    ttl_seconds: int
+    age_seconds: float = 0.0
+    stored_at: datetime
+    expires_at: datetime
+
+
+class CallbackReplyItem(BaseModel):
+    message_id: UUID
+    run_id: UUID
+    project_id: str
+    status: str
+    needs_human: bool
+    action: str | None = None
+    incident_id: str | None = None
+    event_at: datetime | None = None
+    channel_type: str
+    message_type: str
+    send_status: str
+    send_error: str | None = None
+    provider_message_id: str | None = None
+    triggered_by_state: str | None = None
+    content_preview: str | None = None
+    run_state: str | None = None
+    agent_id: str | None = None
+    run_created_at: datetime | None = None
+    callback_reply: dict[str, Any]
+    awooop_status_chain: dict[str, Any] | None = None
+    persisted_awooop_status_chain: dict[str, Any] | None = None
+    km_stale_completion_summary: dict[str, Any] | None = None
+    persisted_km_stale_completion_summary: dict[str, Any] | None = None
+    evidence_capture_status: dict[str, Any] | None = None
+    run_detail_href: str | None = None
+
+
+class OutboundReplyMarkupGapPrefix(BaseModel):
+    prefix: str
+    total: int
+    recent_24h_total: int = 0
+    first_sent_at: datetime | None = None
+    last_sent_at: datetime | None = None
+
+
+class CallbackReplyAuditSummary(BaseModel):
+    schema_version: str
+    project_id: str
+    outbound_total: int
+    outbound_source_envelope_total: int
+    outbound_source_refs_total: int
+    outbound_trace_ref_total: int = 0
+    outbound_incident_ref_total: int
+    outbound_reply_markup_total: int = 0
+    outbound_reply_markup_missing_incident_ref_total: int = 0
+    outbound_reply_markup_missing_incident_ref_recent_1h_total: int = 0
+    outbound_reply_markup_missing_incident_ref_recent_24h_total: int = 0
+    outbound_reply_markup_missing_incident_ref_latest_sent_at: datetime | None = None
+    outbound_reply_markup_missing_trace_ref_total: int = 0
+    outbound_reply_markup_missing_trace_ref_recent_1h_total: int = 0
+    outbound_reply_markup_missing_trace_ref_recent_24h_total: int = 0
+    outbound_reply_markup_missing_trace_ref_latest_sent_at: datetime | None = None
+    outbound_reply_markup_trace_ref_gap_status: str = "clean"
+    outbound_reply_markup_trace_ref_gap_next_action: str = "none"
+    outbound_reply_markup_trace_ref_after_gap_total: int = 0
+    outbound_reply_markup_trace_ref_after_gap_first_sent_at: datetime | None = None
+    outbound_reply_markup_trace_ref_after_gap_latest_sent_at: datetime | None = None
+    outbound_reply_markup_trace_ref_gap_recovery_status: str = "not_needed"
+    outbound_reply_markup_missing_incident_ref_top_prefixes: list[
+        OutboundReplyMarkupGapPrefix
+    ] = Field(default_factory=list)
+    outbound_reply_markup_missing_trace_ref_top_prefixes: list[
+        OutboundReplyMarkupGapPrefix
+    ] = Field(default_factory=list)
+    outbound_failed_total: int
+    callback_total: int
+    callback_sent_total: int
+    callback_fallback_total: int
+    callback_rescue_total: int
+    callback_failed_total: int
+    callback_detail_total: int
+    callback_history_total: int
+    callback_snapshot_captured_total: int
+    callback_snapshot_partial_total: int
+    callback_snapshot_missing_total: int
+    callback_incident_total: int
+    inbound_callback_total: int = 0
+    inbound_callback_recent_24h_total: int = 0
+    inbound_callback_latest_at: datetime | None = None
+    inbound_callback_mirror_status: str = "no_callback_observed"
+    inbound_callback_next_action: str = "press_any_telegram_callback_after_rollout"
+    snapshot_status: str
+    next_action: str
+    latest_outbound_at: datetime | None = None
+    latest_callback_at: datetime | None = None
+
+
+class ListCallbackRepliesResponse(BaseModel):
+    items: list[CallbackReplyItem]
+    total: int
+    page: int
+    per_page: int
+    summary: CallbackReplyAuditSummary | None = None
+    cache: OperatorSummaryCacheInfo | None = None
+
+
+class CicdEventItem(BaseModel):
+    id: str
+    project_id: str
+    alertname: str
+    stage: str | None = None
+    status: str | None = None
+    severity: str | None = None
+    commit_sha: str | None = None
+    triggered_by: str | None = None
+    duration_seconds: int = 0
+    summary: str | None = None
+    description: str | None = None
+    workflow_url: str | None = None
+    alert_id: str | None = None
+    source: str | None = None
+    action_detail: str | None = None
+    needs_attention: bool = False
+    created_at: datetime
+
+
+class ListCicdEventsResponse(BaseModel):
+    items: list[CicdEventItem]
+    total: int
+    limit: int
+
+
+class AiRouteStatusResponse(BaseModel):
+    schema_version: str
+    workload_type: str
+    policy_order: list[dict[str, Any]]
+    selected_provider: str | None = None
+    selected_url: str | None = None
+    selected_model: str | None = None
+    fallback_chain: list[dict[str, Any]]
+    route_reason: str
+    route_source: str
+    route_error: str | None = None
+    health: dict[str, dict[str, Any]]
+    lane_mode: str | None = None
+    active_lane: dict[str, Any] | None = None
+    skipped_lanes: list[dict[str, Any]] = Field(default_factory=list)
+    operator_action: dict[str, Any] | None = None
+    repair_evidence: dict[str, Any] | None = None
+    checked_at: datetime
+
+
 class ApprovalItem(BaseModel):
     run_id: UUID
     project_id: str
     agent_id: str
     created_at: datetime
     timeout_at: datetime | None
+    remediation_summary: dict[str, Any] | None = None
+    awooop_status_chain: dict[str, Any] | None = None
 
 
 class ListApprovalsResponse(BaseModel):
@@ -95,7 +264,8 @@ class DecideApprovalResponse(BaseModel):
     response_model=ListRunsResponse,
     summary="列出 Runs",
     description=(
-        "返回 awooop_run_state 記錄，支援 project_id / state filter 與分頁。\n\n"
+        "返回 awooop_run_state 記錄，支援 project_id / state / remediation_status / "
+        "callback_reply_status / incident_id filter 與分頁。\n\n"
         "- 按 created_at DESC 排序\n"
         "- 注意：此路徑為 /runs/list 以避免與 runs.py 的 /runs/{run_id} 衝突"
     ),
@@ -103,14 +273,102 @@ class DecideApprovalResponse(BaseModel):
 async def list_runs(
     project_id: str | None = Query(None, description="租戶 ID（可選）"),
     state: str | None = Query(None, description="Run 狀態 filter（可選）"),
+    remediation_status: str | None = Query(
+        None,
+        description="AI 證據狀態 filter（no_evidence/mcp_observed/read_only_dry_run/write_observed/blocked/observed）",
+    ),
+    callback_reply_status: str | None = Query(
+        None,
+        description="Telegram callback reply 狀態 filter（no_callback/sent/fallback_sent/rescue_sent/failed/observed）",
+    ),
+    incident_id: str | None = Query(None, description="關聯 Incident ID filter（可選）"),
     page: int = Query(1, ge=1, description="頁碼，從 1 開始"),
     per_page: int = Query(_DEFAULT_PER_PAGE, ge=1, le=_MAX_PER_PAGE, description="每頁筆數"),
 ) -> dict[str, Any]:
     return await list_runs_svc(
-        project_id=project_id, state=state, page=page, per_page=per_page
+        project_id=project_id,
+        state=state,
+        remediation_status=remediation_status,
+        callback_reply_status=callback_reply_status,
+        incident_id=incident_id,
+        page=page,
+        per_page=per_page,
     )
 
 
+@router.get(
+    "/runs/callback-replies",
+    response_model=ListCallbackRepliesResponse,
+    summary="列出 Telegram Callback Reply Evidence",
+    description=(
+        "從 AwoooP outbound mirror 查詢 Telegram 詳情 / 歷史 callback reply 的"
+        "送達、fallback、救援與失敗證據；只讀，不修改 incident、run 或 Telegram 狀態。"
+    ),
+)
+async def list_callback_replies(
+    project_id: str | None = Query(None, description="租戶 ID（可選）"),
+    callback_reply_status: str | None = Query(
+        None,
+        description="Telegram callback reply 狀態 filter（sent/fallback_sent/rescue_sent/failed/observed/no_callback）",
+    ),
+    action: str | None = Query(None, description="Callback action filter（例如 detail/history）"),
+    incident_id: str | None = Query(None, description="關聯 Incident ID filter（可選）"),
+    page: int = Query(1, ge=1, description="頁碼，從 1 開始"),
+    per_page: int = Query(20, ge=1, le=_MAX_PER_PAGE, description="每頁筆數"),
+    refresh: bool = Query(False, description="略過短 TTL 快取並重新聚合"),
+) -> dict[str, Any]:
+    return await list_callback_replies_svc(
+        project_id=project_id,
+        callback_reply_status=callback_reply_status,
+        action=action,
+        incident_id=incident_id,
+        page=page,
+        per_page=per_page,
+        refresh=refresh,
+    )
+
+
+@router.get(
+    "/cicd/events",
+    response_model=ListCicdEventsResponse,
+    summary="列出 CI/CD evidence events",
+    description=(
+        "從 alert_operation_log 讀取 CI/CD notification evidence，供 AwoooP "
+        "Deployments / Run Console 顯示 rollout-risk、success、failed 等階段狀態。"
+    ),
+)
+async def list_cicd_events(
+    project_id: str | None = Query(None, description="租戶 ID（目前支援 awoooi）"),
+    stage: str | None = Query(None, description="CI/CD stage filter（可選）"),
+    status: str | None = Query(None, description="CI/CD status filter（running/success/failed/pending）"),
+    limit: int = Query(12, ge=1, le=50, description="最多返回筆數"),
+) -> dict[str, Any]:
+    return await list_cicd_events_svc(
+        project_id=project_id,
+        stage=stage,
+        status_filter=status,
+        limit=limit,
+    )
+
+
+@router.get(
+    "/ai-route-status",
+    response_model=AiRouteStatusResponse,
+    summary="查詢 AI Provider 路由狀態",
+    description=(
+        "回傳目前 Ollama/Gemini 路由策略、即時 primary、fallback chain 與健康狀態；"
+        "只讀，不觸發推理或自動修復。"
+    ),
+)
+async def get_ai_route_status(
+    workload_type: str | None = Query(
+        "deep_rca",
+        description="工作負載類型，例如 deep_rca/hermes/interactive/embedding/rag/code_review/image_analysis",
+    ),
+) -> dict[str, Any]:
+    return await get_ai_route_status_svc(workload_type=workload_type)
+
+
 @router.get(
     "/runs/{run_id}/detail",
     summary="查詢 Run 詳細時間線",
@@ -126,6 +384,27 @@ async def get_run_detail(
     return await get_run_detail_svc(run_id=run_id, project_id=project_id)
 
 
+@router.get(
+    "/status-chain",
+    summary="查詢 AwoooP 狀態鏈",
+    description=(
+        "依 incident_id 查詢 truth-chain + ADR-100 history 合併後的只讀狀態鏈，"
+        "供 Work Items、Approvals、Monitoring 等操作頁面共用。"
+    ),
+)
+async def get_awooop_status_chain(
+    project_id: str | None = Query(None, description="租戶 ID（可選）"),
+    incident_id: list[str] | None = Query(
+        None,
+        description="Incident ID，可重複傳入以合併同一工作項的多個事件",
+    ),
+) -> dict[str, Any]:
+    return await get_awooop_status_chain_svc(
+        project_id=project_id,
+        incident_ids=incident_id or [],
+    )
+
+
 @router.get(
     "/approvals",
     response_model=ListApprovalsResponse,
@@ -138,8 +417,16 @@ async def get_run_detail(
 async def list_approvals(
     project_id: str | None = Query(None, description="租戶 ID（可選）"),
     run_id: str | None = Query(None, description="Run ID（可選，M8 詳情頁查單筆）"),
+    remediation_status: str | None = Query(
+        None,
+        description="AI 證據狀態 filter（no_evidence/mcp_observed/read_only_dry_run/write_observed/blocked/observed）",
+    ),
 ) -> dict[str, Any]:
-    return await list_approvals_svc(project_id=project_id, run_id=run_id)
+    return await list_approvals_svc(
+        project_id=project_id,
+        run_id=run_id,
+        remediation_status=remediation_status,
+    )
 
 
 @router.post(

apps/api/src/api/v1/platform/truth_chain.py

@@ -0,0 +1,66 @@
+"""AwoooP Operator Console — truth-chain read API."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from fastapi import APIRouter, Depends, Query
+
+from src.core.awooop_operator_auth import (
+    AwoooPOperatorPrincipal,
+    verify_awooop_operator,
+)
+from src.services.awooop_truth_chain_service import (
+    fetch_automation_quality_summary,
+    fetch_truth_chain,
+)
+
+router = APIRouter()
+
+
+@router.get(
+    "/truth-chain/quality/summary",
+    summary="查詢 AI 自動化品質總覽",
+    description=(
+        "T12c read-only aggregate endpoint. 聚合最近 incident 的 automation quality gate，"
+        "讓 Operator 不必逐張 Telegram 卡片判斷是否真正完成 AI 自動修復。"
+        "此總覽不回傳逐筆 examples；source-level truth-chain 詳情仍需 operator auth。"
+    ),
+)
+async def get_automation_quality_summary(
+    project_id: str = Query("awoooi", description="租戶 ID"),
+    hours: int = Query(24, ge=1, le=168, description="回看小時數"),
+    limit: int = Query(200, ge=1, le=500, description="最多評估 incident 數"),
+    refresh: bool = Query(False, description="略過短 TTL 快取並重新聚合"),
+) -> dict[str, Any]:
+    summary = await fetch_automation_quality_summary(
+        project_id=project_id,
+        hours=hours,
+        limit=limit,
+        refresh=refresh,
+    )
+    summary["examples"] = []
+    summary["visibility_note"] = (
+        "Aggregate only. Use /truth-chain/{source_id} with operator auth for source-level details."
+    )
+    return summary
+
+
+@router.get(
+    "/truth-chain/{source_id}",
+    summary="查詢 Telegram / Incident / Drift 真相鏈",
+    description=(
+        "T0 read-only endpoint. 聚合 incident、approval、evidence、MCP、"
+        "automation_operation_log、drift repeat state 與 outbound mirror，"
+        "讓 Operator Console 能判斷 Telegram 卡片目前卡在哪個流程節點。"
+    ),
+)
+async def get_truth_chain(
+    source_id: str,
+    project_id: str = Query("awoooi", description="租戶 ID"),
+    operator: AwoooPOperatorPrincipal = Depends(verify_awooop_operator),
+) -> dict[str, Any]:
+    # operator dependency intentionally gates this read API even though the
+    # principal is not otherwise needed by the aggregation query.
+    _ = operator
+    return await fetch_truth_chain(source_id=source_id, project_id=project_id)

apps/api/src/api/v1/rag.py

@@ -64,6 +64,7 @@ async def rag_debug() -> dict:
     """診斷用：確認容器內 docs 路徑 + Ollama 連線"""
     import os
     from pathlib import Path
+
     import httpx
 
     paths_check = {}
@@ -78,12 +79,23 @@ async def rag_debug() -> dict:
     try:
         async with httpx.AsyncClient(timeout=10.0) as c:
             from src.core.config import get_settings as _gs
+            from src.services.ollama_endpoint_resolver import resolve_ollama_order
+
             settings = _gs()
-            r = await c.post(
-                f"{settings.OLLAMA_URL}/api/embeddings",
-                json={"model": settings.OLLAMA_EMBEDDING_MODEL, "prompt": "test"},
-            )
-            ollama_ok = r.status_code == 200 if r.status_code == 200 else f"http_{r.status_code}"
+            statuses: list[str] = []
+            for endpoint in resolve_ollama_order("embedding"):
+                if not endpoint.url:
+                    continue
+                r = await c.post(
+                    f"{endpoint.url}/api/embeddings",
+                    json={"model": settings.OLLAMA_EMBEDDING_MODEL, "prompt": "test"},
+                )
+                if r.status_code == 200:
+                    ollama_ok = True
+                    break
+                statuses.append(f"{endpoint.provider_name}=http_{r.status_code}")
+            if ollama_ok is not True:
+                ollama_ok = ", ".join(statuses) or "no_endpoint"
     except Exception as e:
         ollama_ok = f"error: {type(e).__name__}: {e}"
 

apps/api/src/api/v1/sentry_webhook.py

@@ -14,12 +14,15 @@ AWOOOI API - Sentry Webhook Handler
 🔴 HARD RULE: 時間顯示使用 Asia/Taipei (UTC+8)
 """
 
+import json
 import uuid
+from typing import Any
 
 import structlog
 from fastapi import APIRouter, BackgroundTasks, HTTPException, Request
 from pydantic import BaseModel
 
+from src.core.awooop_operator_auth import authenticate_awooop_operator_headers
 from src.core.circuit_breaker import get_openclaw_guard
 from src.core.metrics import (
     record_alert_chain_failure,
@@ -35,8 +38,10 @@ from src.models.approval import (
 )
 from src.services.anomaly_counter import get_anomaly_counter
 from src.services.approval_db import get_approval_service
+from src.services.channel_hub import record_external_alert_event
 from src.services.openclaw_http_service import get_openclaw_http_service
 from src.services.sentry_service import get_sentry_service
+
 # 2026-04-27 P3.1-T2 by Claude — Tier-2 三服務感知強化：補 SentryWebhookService 簽章驗證
 from src.services.sentry_webhook_service import (
     SentrySignatureError,
@@ -87,6 +92,114 @@ async def sentry_webhook_health() -> dict:
     return {"status": "ok", "webhook": "sentry"}
 
 
+def _sentry_event_tag(event_data: dict[str, Any], key: str) -> str | None:
+    tags = event_data.get("tags") or []
+    for tag in tags:
+        if isinstance(tag, list | tuple) and len(tag) >= 2 and str(tag[0]) == key:
+            return str(tag[1])
+        if isinstance(tag, dict) and str(tag.get("key")) == key:
+            value = tag.get("value")
+            return str(value) if value is not None else None
+    return None
+
+
+def _is_sentry_upstream_canary(payload: dict[str, Any]) -> bool:
+    data = payload.get("data") if isinstance(payload, dict) else None
+    if not isinstance(data, dict) or payload.get("action") != "triggered":
+        return False
+    issue_data = data.get("issue") if isinstance(data.get("issue"), dict) else {}
+    event_data = data.get("event") if isinstance(data.get("event"), dict) else {}
+    issue_id = str(issue_data.get("id") or "")
+    short_id = str(issue_data.get("shortId") or "")
+    title = str(issue_data.get("title") or "")
+    return (
+        issue_id.startswith("awoooi-canary-")
+        or short_id.upper().startswith("AWOOOI-CANARY")
+        or title == "AwoooPSourceProviderCanary"
+        or (_sentry_event_tag(event_data, "awoooi_canary") or "").lower() == "true"
+    )
+
+
+async def _record_sentry_upstream_canary(
+    payload: dict[str, Any],
+    request: Request,
+) -> dict[str, Any]:
+    operator = authenticate_awooop_operator_headers(
+        request.headers.get("x-awooop-operator-id"),
+        request.headers.get("x-awooop-operator-key"),
+    )
+    data = payload.get("data") if isinstance(payload.get("data"), dict) else {}
+    issue_data = data.get("issue") if isinstance(data.get("issue"), dict) else {}
+    event_data = data.get("event") if isinstance(data.get("event"), dict) else {}
+    issue_id = str(
+        issue_data.get("id")
+        or issue_data.get("shortId")
+        or _sentry_event_tag(event_data, "run_ref")
+        or "awoooi-canary-unknown"
+    )
+    source_url = (
+        issue_data.get("permalink")
+        or issue_data.get("web_url")
+        or issue_data.get("url")
+    )
+    event_uuid = await record_external_alert_event(
+        project_id="awoooi",
+        provider="sentry",
+        event_id=issue_id,
+        stage="upstream_canary",
+        title=str(issue_data.get("title") or "AwoooPSourceProviderCanary"),
+        severity=str(issue_data.get("level") or "info"),
+        namespace="awoooi-prod",
+        target_resource=str(issue_data.get("culprit") or "source-provider-ingestion"),
+        fingerprint=f"source-provider-canary:sentry:{issue_id}",
+        source_url=source_url,
+        labels={
+            "project": issue_data.get("project", {}),
+            "level": issue_data.get("level", "info"),
+            "awoooi_canary": "true",
+            "operator_id": operator.operator_id,
+            "telegram": "not_sent",
+            "incident": "not_created",
+            "approval": "not_created",
+        },
+        annotations={
+            "message": event_data.get("message"),
+            "summary": (
+                "Operator-signed Sentry webhook canary; records upstream "
+                "source evidence without creating incident, approval, or Telegram."
+            ),
+        },
+        payload={
+            "raw_canary": payload,
+            "operator_id": operator.operator_id,
+            "auth_method": operator.auth_method,
+            "side_effects": {
+                "incident_created": False,
+                "approval_created": False,
+                "telegram_sent": False,
+                "openclaw_called": False,
+            },
+        },
+    )
+    if event_uuid is None:
+        raise HTTPException(
+            status_code=500,
+            detail="sentry upstream canary was not recorded",
+        )
+    return {
+        "status": "canary_recorded",
+        "provider": "sentry",
+        "event_id": issue_id,
+        "conversation_event_id": str(event_uuid),
+        "side_effects": {
+            "incident_created": False,
+            "approval_created": False,
+            "telegram_sent": False,
+            "openclaw_called": False,
+        },
+    }
+
+
 @router.post("/error")
 async def handle_sentry_error(
     request: Request,
@@ -108,6 +221,14 @@ async def handle_sentry_error(
     try:
         # 2026-04-27 P3.1-T2 by Claude — Tier-2 三服務感知強化：接入 SentryWebhookService 簽章驗證
         body = await request.body()
+        try:
+            payload_from_body = json.loads(body.decode("utf-8") or "{}")
+        except json.JSONDecodeError:
+            payload_from_body = {}
+
+        if isinstance(payload_from_body, dict) and _is_sentry_upstream_canary(payload_from_body):
+            return await _record_sentry_upstream_canary(payload_from_body, request)
+
         sig_header = request.headers.get("sentry-hook-signature", "")
         try:
             verify_sentry_signature(body, sig_header)
@@ -124,16 +245,60 @@ async def handle_sentry_error(
 
         # 提取錯誤資訊
         issue_data = payload.get("data", {}).get("issue", {})
+        event_data = payload.get("data", {}).get("event", {})
+        issue_id = issue_data.get("id")
+        source_url = (
+            issue_data.get("permalink")
+            or issue_data.get("web_url")
+            or issue_data.get("url")
+        )
+
+        background_tasks.add_task(
+            record_external_alert_event,
+            project_id="awoooi",
+            provider="sentry",
+            event_id=str(issue_id or issue_data.get("shortId") or "unknown"),
+            stage="received",
+            title=str(issue_data.get("title") or "Sentry issue"),
+            severity=str(issue_data.get("level") or "error"),
+            namespace="sentry",
+            target_resource=str(issue_data.get("culprit") or issue_data.get("project", {}).get("slug") or "unknown"),
+            fingerprint=f"sentry-{issue_id or issue_data.get('shortId') or 'unknown'}",
+            source_url=source_url,
+            labels={
+                "project": issue_data.get("project", {}),
+                "level": issue_data.get("level"),
+                "culprit": issue_data.get("culprit"),
+            },
+            annotations={"message": event_data.get("message")},
+            payload=payload,
+        )
 
         # Phase 10.2.1: 去重檢查 (10 分鐘內不重複發送)
-        issue_id = issue_data.get("id")
         sentry_service = get_sentry_service()
         if not await sentry_service.check_dedup(issue_id, ttl=SENTRY_DEDUP_TTL):
+            background_tasks.add_task(
+                record_external_alert_event,
+                project_id="awoooi",
+                provider="sentry",
+                event_id=str(issue_id or issue_data.get("shortId") or "unknown"),
+                stage="deduplicated",
+                title=str(issue_data.get("title") or "Sentry issue"),
+                severity=str(issue_data.get("level") or "error"),
+                namespace="sentry",
+                target_resource=str(issue_data.get("culprit") or issue_data.get("project", {}).get("slug") or "unknown"),
+                fingerprint=f"sentry-{issue_id or issue_data.get('shortId') or 'unknown'}",
+                source_url=source_url,
+                labels={"project": issue_data.get("project", {}), "level": issue_data.get("level")},
+                annotations={"message": event_data.get("message")},
+                payload={"dedup_ttl": SENTRY_DEDUP_TTL},
+                is_duplicate=True,
+            )
             return {"status": "deduplicated", "issue_id": issue_id, "ttl": SENTRY_DEDUP_TTL}
-        event_data = payload.get("data", {}).get("event", {})
 
         error_context = {
             "issue_id": issue_data.get("id"),
+            "source_url": source_url,
             "title": issue_data.get("title"),
             "culprit": issue_data.get("culprit"),
             "level": issue_data.get("level"),
@@ -169,6 +334,8 @@ async def handle_sentry_error(
             "message": "Analysis scheduled"
         }
 
+    except HTTPException:
+        raise
     except Exception as e:
         logger.exception("Sentry webhook processing failed")
         raise HTTPException(status_code=500, detail=str(e)) from e
@@ -256,6 +423,29 @@ async def analyze_and_comment(
             analysis=analysis,
             anomaly_frequency=frequency_dict,
         )
+        await record_external_alert_event(
+            project_id="awoooi",
+            provider="sentry",
+            event_id=str(issue_id or error_context.get("issue_id") or "unknown"),
+            stage="approval_linked",
+            title=str(error_context.get("title") or "Sentry issue"),
+            severity=str(error_context.get("level") or "error"),
+            namespace="sentry",
+            target_resource=str(error_context.get("culprit") or error_context.get("project") or "unknown"),
+            fingerprint=f"sentry-{issue_id or error_context.get('issue_id') or 'unknown'}",
+            approval_id=approval_id,
+            source_url=error_context.get("source_url"),
+            labels={
+                "project": error_context.get("project"),
+                "level": error_context.get("level"),
+            },
+            annotations={"message": error_context.get("message")},
+            payload={
+                "anomaly_frequency": frequency_dict,
+                "ai_analyzed": analysis is not None,
+                "ai_provider": analysis.analyzed_by if analysis else None,
+            },
+        )
 
         # 4. 發送 Telegram 告警 (含頻率資訊)
         await send_sentry_telegram_alert(

apps/api/src/api/v1/signoz_webhook.py

@@ -1,7 +1,3 @@
-from __future__ import annotations
-
-import asyncio
-
 """
 AWOOOI API - SignOz Webhook Handler
 ====================================
@@ -17,12 +13,17 @@ AWOOOI API - SignOz Webhook Handler
 🔴 HARD RULE: 時間顯示使用 Asia/Taipei (UTC+8)
 """
 
+from __future__ import annotations
+
+import asyncio
 import uuid
+from typing import TYPE_CHECKING
 
 import structlog
 from fastapi import APIRouter, BackgroundTasks, HTTPException, Request
 from pydantic import BaseModel
 
+from src.core.awooop_operator_auth import authenticate_awooop_operator_headers
 from src.core.metrics import (
     record_alert_chain_failure,
     record_alert_chain_success,
@@ -37,10 +38,14 @@ from src.models.approval import (
 )
 from src.services.anomaly_counter import get_anomaly_counter
 from src.services.approval_db import get_approval_service
+from src.services.channel_hub import record_external_alert_event
 from src.services.incident_service import get_incident_service
 from src.services.telegram_gateway import get_telegram_gateway
 from src.utils.timezone import now_taipei_iso
 
+if TYPE_CHECKING:
+    from src.services.openclaw import LLMAnalysisResult
+
 logger = structlog.get_logger(__name__)
 
 router = APIRouter(prefix="/webhooks/signoz", tags=["SignOz Webhook"])
@@ -67,6 +72,101 @@ class SignOzAlertPayload(BaseModel):
     generatorURL: str | None = None
 
 
+def _is_signoz_upstream_canary(alert: dict) -> bool:
+    labels = alert.get("labels", {}) if isinstance(alert.get("labels"), dict) else {}
+    annotations = (
+        alert.get("annotations", {})
+        if isinstance(alert.get("annotations"), dict)
+        else {}
+    )
+    alert_name = str(alert.get("alertname") or labels.get("alertname") or "")
+    return (
+        str(labels.get("awoooi_canary", "")).lower() == "true"
+        or alert_name == "AwoooPSourceProviderCanary"
+        or str(annotations.get("awooop_canary", "")).lower() == "true"
+    )
+
+
+async def _record_signoz_upstream_canary(
+    alert: dict,
+    request: Request,
+) -> dict:
+    operator = authenticate_awooop_operator_headers(
+        request.headers.get("x-awooop-operator-id"),
+        request.headers.get("x-awooop-operator-key"),
+    )
+    labels = alert.get("labels", {}) if isinstance(alert.get("labels"), dict) else {}
+    annotations = (
+        alert.get("annotations", {})
+        if isinstance(alert.get("annotations"), dict)
+        else {}
+    )
+    alert_name = str(alert.get("alertname") or labels.get("alertname") or "AwoooPSourceProviderCanary")
+    run_ref = str(labels.get("run_ref") or labels.get("fingerprint") or "unknown")
+    event_id = f"awooop-canary-{run_ref}"
+    severity = str(labels.get("severity") or "info")
+    service_name = str(labels.get("service_name") or labels.get("service") or "source-provider-ingestion")
+    namespace = str(labels.get("namespace") or "awoooi-prod")
+    fingerprint = str(labels.get("fingerprint") or f"source-provider-canary:signoz:{run_ref}")
+    event_uuid = await record_external_alert_event(
+        project_id="awoooi",
+        provider="signoz",
+        event_id=event_id,
+        stage="upstream_canary",
+        title=alert_name,
+        severity=severity,
+        namespace=namespace,
+        target_resource=service_name,
+        fingerprint=fingerprint,
+        source_url=alert.get("generatorURL"),
+        labels={
+            **labels,
+            "awoooi_canary": "true",
+            "operator_id": operator.operator_id,
+            "telegram": "not_sent",
+            "incident": "not_created",
+            "approval": "not_created",
+        },
+        annotations={
+            **annotations,
+            "summary": annotations.get("summary")
+            or (
+                "Operator-signed SignOz webhook canary; records upstream "
+                "source evidence without creating incident, approval, or Telegram."
+            ),
+        },
+        payload={
+            "raw_canary": alert,
+            "operator_id": operator.operator_id,
+            "auth_method": operator.auth_method,
+            "side_effects": {
+                "incident_created": False,
+                "approval_created": False,
+                "telegram_sent": False,
+                "openclaw_called": False,
+            },
+        },
+    )
+    if event_uuid is None:
+        raise HTTPException(
+            status_code=500,
+            detail="signoz upstream canary was not recorded",
+        )
+    return {
+        "status": "canary_recorded",
+        "provider": "signoz",
+        "event_id": event_id,
+        "alert_name": alert_name,
+        "conversation_event_id": str(event_uuid),
+        "side_effects": {
+            "incident_created": False,
+            "approval_created": False,
+            "telegram_sent": False,
+            "openclaw_called": False,
+        },
+    }
+
+
 @router.post("/alert")
 async def handle_signoz_alert(
     request: Request,
@@ -99,11 +199,35 @@ async def handle_signoz_alert(
                 results.append({"status": "ignored", "reason": "not firing"})
                 continue
 
+            if _is_signoz_upstream_canary(alert):
+                results.append(await _record_signoz_upstream_canary(alert, request))
+                continue
+
             # 提取告警資訊
             alert_name = alert.get("alertname", alert.get("labels", {}).get("alertname", "unknown"))
             labels = alert.get("labels", {})
             annotations = alert.get("annotations", {})
             severity = labels.get("severity", "warning")
+            source_url = alert.get("generatorURL")
+            service_name = labels.get("service_name", labels.get("service", "unknown"))
+            fingerprint = labels.get("fingerprint") or f"signoz-{alert_name}-{service_name}"
+
+            background_tasks.add_task(
+                record_external_alert_event,
+                project_id="awoooi",
+                provider="signoz",
+                event_id=str(fingerprint),
+                stage="received",
+                title=str(alert_name),
+                severity=str(severity),
+                namespace=str(labels.get("namespace", "signoz")),
+                target_resource=str(service_name),
+                fingerprint=str(fingerprint),
+                source_url=source_url,
+                labels=labels,
+                annotations=annotations,
+                payload=alert,
+            )
 
             # 背景處理
             background_tasks.add_task(
@@ -113,6 +237,8 @@ async def handle_signoz_alert(
                 annotations=annotations,
                 severity=severity,
                 starts_at=alert.get("startsAt"),
+                source_url=source_url,
+                raw_payload=alert,
             )
 
             results.append({
@@ -122,6 +248,8 @@ async def handle_signoz_alert(
 
         return {"status": "ok", "processed": len(results), "results": results}
 
+    except HTTPException:
+        raise
     except Exception as e:
         logger.exception("signoz_webhook_error", error=str(e))
         raise HTTPException(status_code=500, detail=str(e)) from e
@@ -133,6 +261,8 @@ async def process_signoz_alert(
     annotations: dict,
     severity: str,
     starts_at: str | None,
+    source_url: str | None = None,
+    raw_payload: dict | None = None,
 ):
     """
     背景處理 SignOz 告警
@@ -190,6 +320,7 @@ async def process_signoz_alert(
             "annotations": annotations,
             "fingerprint": f"signoz-{alert_name}-{labels.get('service_name', 'unknown')}",
         }
+        fingerprint = signal_data["fingerprint"]
         # ADR-037: 傳遞頻率統計到 Incident
         incident = await incident_service.create_incident_from_signal(
             signal_data, frequency_stats=anomaly_frequency
@@ -229,6 +360,30 @@ async def process_signoz_alert(
             anomaly_frequency=anomaly_frequency,
             analysis_result=analysis_result,  # 帶入 AI 結果
         )
+        await record_external_alert_event(
+            project_id="awoooi",
+            provider="signoz",
+            event_id=str(fingerprint),
+            stage="incident_linked",
+            title=str(alert_name),
+            severity=str(severity),
+            namespace=str(labels.get("namespace", "signoz")),
+            target_resource=str(labels.get("service_name", labels.get("service", "unknown"))),
+            fingerprint=str(fingerprint),
+            incident_id=str(incident.incident_id),
+            approval_id=str(approval_id),
+            source_url=source_url or trace_url,
+            labels=labels,
+            annotations=annotations,
+            payload={
+                "raw_alert": raw_payload or {},
+                "trace_url": trace_url,
+                "has_signoz_metrics": bool(signoz_metrics),
+                "ai_provider": ai_provider,
+                "tokens": tokens,
+                "cost": cost,
+            },
+        )
 
         # =================================================================
         # Step 5: 發送 Telegram 告警
@@ -282,7 +437,7 @@ async def create_signoz_approval(
     severity: str,
     incident_id: str,
     anomaly_frequency: dict | None = None,
-    analysis_result: "LLMAnalysisResult" | None = None,
+    analysis_result: LLMAnalysisResult | None = None,
 ) -> str:
     """
     為 SignOz 告警建立 Approval 記錄
@@ -379,7 +534,7 @@ async def send_signoz_telegram(
     annotations: dict,
     severity: str,
     anomaly_frequency: dict | None = None,
-    analysis_result: "LLMAnalysisResult" | None = None,
+    analysis_result: LLMAnalysisResult | None = None,
     ai_provider: str = "none",
 ):
     """
@@ -442,6 +597,7 @@ async def _send_log_summary_notification(
     帶 5s 軟超時：超時後摘要繼續生成並存 Redis，不阻塞告警主流程
     """
     import html as _html
+
     from src.services.log_summary_service import get_log_summary_service
     from src.services.telegram_gateway import get_telegram_gateway
 

apps/api/src/api/v1/telegram.py

@@ -19,6 +19,7 @@ Endpoints:
 - 每個 Nonce 只能使用一次
 """
 
+import asyncio
 from uuid import UUID
 
 from fastapi import APIRouter, HTTPException, status
@@ -27,6 +28,8 @@ from pydantic import BaseModel
 from src.core.config import settings
 from src.core.logging import get_logger
 from src.services.approval_db import get_approval_service
+from src.services.approval_execution import get_execution_service
+from src.services.incident_approval_service import get_incident_approval_service
 from src.services.security_interceptor import (
     NonceReplayError,
     UserNotWhitelistedError,
@@ -64,6 +67,80 @@ class TestPushRequest(BaseModel):
     incident_id: str = ""
 
 
+async def _run_telegram_approved_execution(approval) -> None:
+    """Run the approved action that originated from a Telegram callback."""
+    approval_id = str(getattr(approval, "id", ""))
+    incident_id = getattr(approval, "incident_id", None)
+    try:
+        result = await get_execution_service().execute_approved_action(approval)
+        logger.info(
+            "telegram_approval_execution_completed",
+            approval_id=approval_id,
+            incident_id=incident_id,
+            success=bool(result),
+        )
+    except Exception as exc:
+        logger.error(
+            "telegram_approval_execution_failed",
+            approval_id=approval_id,
+            incident_id=incident_id,
+            error=str(exc),
+        )
+
+
+def _schedule_telegram_approved_execution(approval) -> bool:
+    """Schedule execution after Telegram approval reaches required signatures."""
+    try:
+        asyncio.create_task(_run_telegram_approved_execution(approval))
+        logger.info(
+            "telegram_approval_execution_scheduled",
+            approval_id=str(getattr(approval, "id", "")),
+            incident_id=getattr(approval, "incident_id", None),
+        )
+        return True
+    except Exception as exc:
+        logger.error(
+            "telegram_approval_execution_schedule_failed",
+            approval_id=str(getattr(approval, "id", "")),
+            incident_id=getattr(approval, "incident_id", None),
+            error=str(exc),
+        )
+        return False
+
+
+async def _finalize_telegram_approval(approval, execution_triggered: bool) -> bool:
+    """Complete the execution handoff for Telegram approvals.
+
+    ApprovalDBService only records the signature/status transition. The actual
+    executor scheduling lives in API callers, so Telegram must mirror the REST
+    approval endpoint instead of stopping at a visual approval stamp.
+    """
+    if not execution_triggered:
+        return False
+    return _schedule_telegram_approved_execution(approval)
+
+
+async def _sync_telegram_rejection(approval_id: str) -> bool:
+    """Keep Incident state aligned when an approval is rejected from Telegram."""
+    try:
+        await get_incident_approval_service().on_approval_status_change(
+            approval_id=approval_id,
+            new_status="rejected",
+        )
+        logger.info(
+            "telegram_rejection_incident_synced",
+            approval_id=approval_id,
+        )
+        return True
+    except Exception as exc:
+        logger.error(
+            "telegram_rejection_incident_sync_failed",
+            approval_id=approval_id,
+            error=str(exc),
+        )
+        return False
+
+
 # =============================================================================
 # Endpoints
 # =============================================================================
@@ -139,6 +216,17 @@ async def telegram_webhook(
     # =========================================================================
     try:
         gateway = get_telegram_gateway()
+        mirror_callback = getattr(gateway, "mirror_callback_query_received", None)
+        if callable(mirror_callback):
+            await mirror_callback(
+                update_id=update.update_id,
+                callback_query_id=callback_query_id,
+                callback_data=callback_data,
+                user_id=user_id,
+                username=username,
+                message_id=message_id,
+                chat_id=message.get("chat", {}).get("id"),
+            )
         result = await gateway.handle_callback(
             callback_query_id=callback_query_id,
             callback_data=callback_data,
@@ -198,21 +286,50 @@ async def telegram_webhook(
             )
 
             if approval:
+                status_value = approval.status.value if hasattr(approval.status, "value") else str(approval.status)
+                if (
+                    "Cannot sign" in msg
+                    or "already signed" in msg
+                    or "Concurrent modification" in msg
+                ):
+                    logger.info(
+                        "telegram_approval_ignored_already_processed",
+                        approval_id=approval_id,
+                        user_id=user_id,
+                        status=status_value,
+                        message=msg,
+                    )
+                    await _log_user_action("approve_duplicate", False, getattr(approval, "incident_id", None))
+                    return {
+                        "ok": True,
+                        "message": "Already processed",
+                        "approval_id": approval_id,
+                        "status": status_value,
+                        "execution_triggered": False,
+                        "execution_scheduled": False,
+                    }
+
+                execution_scheduled = await _finalize_telegram_approval(
+                    approval=approval,
+                    execution_triggered=execution_triggered,
+                )
                 logger.info(
                     "telegram_approval_signed",
                     approval_id=approval_id,
                     user_id=user_id,
-                    status=approval.status.value,
+                    status=status_value,
                     execution_triggered=execution_triggered,
+                    execution_scheduled=execution_scheduled,
                 )
                 await _log_user_action("approve", True, getattr(approval, "incident_id", None))
 
                 return {
                     "ok": True,
-                    "message": "Approved",
+                    "message": "Approved" if execution_triggered else "Signed",
                     "approval_id": approval_id,
-                    "status": approval.status.value,
+                    "status": status_value,
                     "execution_triggered": execution_triggered,
+                    "execution_scheduled": execution_scheduled,
                 }
 
         elif action == "reject":
@@ -224,10 +341,12 @@ async def telegram_webhook(
             )
 
             if approval:
+                incident_synced = await _sync_telegram_rejection(approval_id)
                 logger.info(
                     "telegram_approval_rejected",
                     approval_id=approval_id,
                     user_id=user_id,
+                    incident_synced=incident_synced,
                 )
                 await _log_user_action("reject", False, getattr(approval, "incident_id", None))
 
@@ -236,6 +355,7 @@ async def telegram_webhook(
                     "message": "Rejected",
                     "approval_id": approval_id,
                     "status": approval.status.value,
+                    "incident_synced": incident_synced,
                 }
 
         return {"ok": False, "message": "Unknown action"}

apps/api/src/api/v1/telegram_webhook.py

@@ -71,6 +71,29 @@ async def telegram_webhook(request: Request) -> dict:
         update_id=body.get("update_id"),
     )
 
+    if update_type == "callback_query":
+        callback = body.get("callback_query", {}) or {}
+        message = callback.get("message", {}) or {}
+        user = callback.get("from", {}) or {}
+        callback_query_id = callback.get("id")
+        callback_data = callback.get("data")
+        user_id = user.get("id")
+        if callback_query_id and callback_data and user_id:
+            from src.services.telegram_gateway import get_telegram_gateway
+
+            gateway = get_telegram_gateway()
+            mirror_callback = getattr(gateway, "mirror_callback_query_received", None)
+            if callable(mirror_callback):
+                await mirror_callback(
+                    update_id=body.get("update_id"),
+                    callback_query_id=callback_query_id,
+                    callback_data=callback_data,
+                    user_id=user_id,
+                    username=user.get("username") or user.get("first_name") or str(user_id),
+                    message_id=message.get("message_id"),
+                    chat_id=(message.get("chat") or {}).get("id"),
+                )
+
     # WS5: chat_member 同步 Approvers 白名單（ADR-093）
     if update_type in ("chat_member", "my_chat_member") or (
         "chat_member" in body or "my_chat_member" in body

apps/api/src/api/v1/webhooks.py

@@ -55,7 +55,10 @@ from src.services.alertmanager_llm_guard import (
 from src.services.approval_db import get_approval_service
 from src.services.auto_approve import get_auto_approve_policy
 from src.services.auto_repair_service import AutoRepairService
-from src.services.channel_hub import record_grouped_alert_event
+from src.services.channel_hub import (
+    record_alertmanager_event,
+    record_grouped_alert_event,
+)
 
 # Phase 15.2: Trace Context (moved to SignalProducerService)
 # get_trace_context 已移至 Service 層
@@ -136,6 +139,38 @@ def _should_use_alertmanager_rule_first(
     )
 
 
+async def _analyze_alertmanager_with_timeout(
+    openclaw,
+    alert_context: dict,
+    *,
+    alert_id: str,
+    alertname: str,
+) -> tuple:
+    """Run Alertmanager AI analysis without letting it block the workflow forever."""
+
+    try:
+        return await asyncio.wait_for(
+            openclaw.analyze_alert(alert_context),
+            timeout=ALERTMANAGER_BACKGROUND_AI_TIMEOUT_SECONDS,
+        )
+    except TimeoutError:
+        logger.warning(
+            "alertmanager_openclaw_timeout_fallback",
+            alert_id=alert_id,
+            alertname=alertname,
+            timeout_sec=ALERTMANAGER_BACKGROUND_AI_TIMEOUT_SECONDS,
+        )
+        return None, "fallback_timeout", "", None, "", 0, 0.0
+    except Exception as exc:
+        logger.warning(
+            "alertmanager_openclaw_failed_fallback",
+            alert_id=alert_id,
+            alertname=alertname,
+            error=str(exc),
+        )
+        return None, "fallback_error", "", None, "", 0, 0.0
+
+
 async def _escalate_auto_repair_unavailable(
     *,
     incident_id: str,
@@ -163,6 +198,19 @@ async def _escalate_auto_repair_unavailable(
     )
 
 
+def _auto_repair_action_label(result, fallback_target: str) -> str:
+    """Build a verifier label that includes the actual playbook steps."""
+    playbook_id = getattr(result, "playbook_id", None) or "unknown"
+    steps = getattr(result, "executed_steps", None) or []
+    step_text = " | ".join(str(step) for step in steps).strip()
+    if not step_text:
+        step_text = fallback_target
+    step_text = " ".join(step_text.split())
+    if len(step_text) > 240:
+        step_text = f"{step_text[:237]}..."
+    return f"auto_repair_playbook:{playbook_id} {step_text}".strip()
+
+
 async def _try_auto_repair_background(
     incident_id: str,
     approval_id: str,
@@ -252,6 +300,46 @@ async def _try_auto_repair_background(
             },
         )
 
+        _pre_execution_snapshot = None
+        try:
+            from src.core.feature_flags import aiops_flags
+
+            if aiops_flags.is_sub_flag_enabled("AIOPS_P1_PRE_DECISION_INVESTIGATOR"):
+                from src.services.evidence_snapshot import get_latest_snapshot
+                from src.services.post_execution_verifier import get_post_execution_verifier
+
+                _pre_execution_snapshot = await get_latest_snapshot(incident_id)
+                if _pre_execution_snapshot is None:
+                    from src.services.pre_decision_investigator import (
+                        get_pre_decision_investigator,
+                    )
+
+                    _pre_execution_snapshot = await asyncio.wait_for(
+                        get_pre_decision_investigator().investigate(incident),
+                        timeout=60.0,
+                    )
+                if _pre_execution_snapshot is not None:
+                    await asyncio.wait_for(
+                        get_post_execution_verifier().capture_pre_execution_state(
+                            incident,
+                            _pre_execution_snapshot,
+                        ),
+                        timeout=30.0,
+                    )
+        except asyncio.TimeoutError:
+            logger.warning(
+                "auto_repair_pre_state_capture_timeout",
+                incident_id=incident_id,
+                approval_id=approval_id,
+            )
+        except Exception as _pre_state_err:
+            logger.warning(
+                "auto_repair_pre_state_capture_failed",
+                incident_id=incident_id,
+                approval_id=approval_id,
+                error=str(_pre_state_err),
+            )
+
         # 執行自動修復
         logger.info(
             "auto_repair_executing",
@@ -263,6 +351,7 @@ async def _try_auto_repair_background(
             playbook=decision.playbook,
             is_cold_start=decision.is_cold_start,
             similarity_score=decision.similarity_score,
+            run_post_verification=False,
         )
 
         logger.info(
@@ -273,6 +362,20 @@ async def _try_auto_repair_background(
 
         # 記錄執行結果
         if result:
+            try:
+                await get_approval_service().update_execution_status(
+                    approval_id=approval_id,
+                    success=result.success,
+                    error_message=result.error,
+                )
+            except Exception as _approval_status_err:
+                logger.warning(
+                    "auto_repair_approval_status_update_failed",
+                    approval_id=approval_id,
+                    incident_id=incident_id,
+                    error=str(_approval_status_err),
+                )
+
             await op_log.append(
                 "EXECUTION_COMPLETED",
                 incident_id=incident_id,
@@ -336,11 +439,10 @@ async def _try_auto_repair_background(
                     from src.services.evidence_snapshot import get_latest_snapshot
                     from src.services.learning_service import get_learning_service
 
-                    _snapshot = await get_latest_snapshot(incident_id)
-                    _action_label = (
-                        f"{target_resource}:{namespace}"
-                        if not result.success
-                        else f"auto_repair_playbook:{result.playbook_id}"
+                    _snapshot = _pre_execution_snapshot or await get_latest_snapshot(incident_id)
+                    _action_label = _auto_repair_action_label(
+                        result,
+                        fallback_target=f"{target_resource}:{namespace}",
                     )
                     _verifier = get_post_execution_verifier()
                     _verify_result = await asyncio.wait_for(
@@ -792,6 +894,7 @@ async def verify_webhook_signature(
 
 # 戰略 B: 滑動時間窗 (ADR-073: 5 分鐘改 30 分鐘，防同一問題反覆重建 Incident，2026-04-12 ogt)
 DEBOUNCE_WINDOW_MINUTES = 30
+ALERTMANAGER_BACKGROUND_AI_TIMEOUT_SECONDS = 90.0
 
 
 # =============================================================================
@@ -1105,7 +1208,12 @@ async def receive_alert(
         # 呼叫 OpenClaw LLM 分析 (v7.0 含 SignOz 整合)
         # 2026-03-29 ogt: 加入 Token/Cost 追蹤
         openclaw = get_openclaw()
-        analysis_result, ai_provider, raw_response, signoz_metrics, signoz_trace_url, ai_tokens, ai_cost = await openclaw.analyze_alert(alert_context)
+        analysis_result, ai_provider, raw_response, signoz_metrics, signoz_trace_url, ai_tokens, ai_cost = await _analyze_alertmanager_with_timeout(
+            openclaw,
+            alert_context,
+            alert_id=alert_id,
+            alertname=alert.alert_type,
+        )
 
         if analysis_result:
             # LLM 分析成功
@@ -1439,6 +1547,39 @@ class AlertmanagerPayload(BaseModel):
     alerts: list[AlertmanagerAlert]
 
 
+_CICD_JOB_STATUSES = frozenset({"running", "success", "failed", "pending"})
+
+
+def _cicd_job_status_from_alert(alert: AlertmanagerAlert) -> str:
+    """將 CI/CD Alertmanager label 轉成 TelegramGateway 支援的狀態。
+
+    2026-05-12 Codex: Gitea workflow 先送進 AWOOI API，不能只靠
+    severity=info 推 success，否則 failed/pending 事件進 AwoooP 後語義會失真。
+    """
+    labels = alert.labels or {}
+    for key in ("status", "job_status", "ci_status"):
+        value = str(labels.get(key) or "").strip().lower()
+        if value in _CICD_JOB_STATUSES:
+            return value
+
+    severity = str(labels.get("severity") or "").strip().lower()
+    if severity == "info":
+        return "success"
+    if severity in {"critical", "error"}:
+        return "failed"
+    return "running"
+
+
+def _cicd_duration_seconds_from_alert(alert: AlertmanagerAlert) -> int:
+    labels = alert.labels or {}
+    raw = labels.get("duration_seconds") or labels.get("duration") or 0
+    try:
+        value = int(str(raw).strip())
+    except (TypeError, ValueError):
+        return 0
+    return max(value, 0)
+
+
 def is_internal_ip(client_ip: str) -> bool:
     """檢查是否為內網 IP"""
     import ipaddress
@@ -1475,6 +1616,11 @@ async def _process_new_alert_background(
     try:
         service = get_approval_service()
         openclaw = get_openclaw()
+        traced_alert_labels = {
+            **(alert_labels or {}),
+            "fingerprint": fingerprint,
+            "alert_id": alert_id,
+        }
 
         rule_response = match_rule(alert_context)
         should_bypass_llm = _should_use_alertmanager_rule_first(rule_response, alert_category)
@@ -1602,6 +1748,10 @@ async def _process_new_alert_background(
             # 2026-04-27 ogt + Claude Sonnet 4.6: CS2 規則引擎自動執行
             # 設計：is_rule_based=True 確定性高，滿足條件直接執行，不等人工審核
             # 安全防線：CRITICAL / destructive patterns / NO_ACTION / 空 kubectl → 全部降級 PENDING
+            _cs2_auto_approval = None
+            _cs2_executor = None
+            _cs2_exec_success: bool | None = None
+            _cs2_exec_error: str | None = None
             try:
                 from src.models.approval import ApprovalRequest, ApprovalStatus
                 from src.services.approval_execution import ApprovalExecutionService
@@ -1625,6 +1775,7 @@ async def _process_new_alert_background(
                     )
                     # 使用 DB 中剛建立的 approval.id 讓 executor 可回寫
                     _auto_approval.id = approval.id
+                    _cs2_auto_approval = _auto_approval
 
                     _cs2_executor = ApprovalExecutionService()
                     _cs2_exec_success = await _cs2_executor.execute_approved_action(_auto_approval)
@@ -1647,6 +1798,8 @@ async def _process_new_alert_background(
                         exec_success=_cs2_exec_success,
                     )
             except Exception as _auto_err:
+                _cs2_exec_success = False if _cs2_auto_approval is not None else None
+                _cs2_exec_error = str(_auto_err)
                 logger.warning(
                     "cs2_auto_execute_failed_degraded_to_pending",
                     approval_id=str(approval.id),
@@ -1662,7 +1815,7 @@ async def _process_new_alert_background(
                 message=message,
                 source="alertmanager",
                 alertname=alertname,
-                alert_labels=alert_labels,
+                alert_labels=traced_alert_labels,
                 notification_type=notification_type,
                 alert_category=alert_category,
             )
@@ -1678,6 +1831,41 @@ async def _process_new_alert_background(
                     error=str(_meta_err),
                 )
 
+            await record_alertmanager_event(
+                project_id="awoooi",
+                alert_id=alert_id,
+                alertname=alertname,
+                severity=severity,
+                namespace=namespace,
+                target_resource=target_resource,
+                fingerprint=fingerprint,
+                stage="incident_linked",
+                notification_type=notification_type,
+                alert_category=alert_category,
+                incident_id=incident_id,
+                approval_id=str(approval.id),
+                repeat_count=1,
+                labels=traced_alert_labels,
+                annotations=alert_context.get("annotations", {}),
+            )
+
+            if _cs2_auto_approval is not None and _cs2_exec_success is not None:
+                try:
+                    _cs2_auto_approval.incident_id = incident_id
+                    _cs2_executor = _cs2_executor or ApprovalExecutionService()
+                    await _cs2_executor.finalize_auto_approved_execution(
+                        _cs2_auto_approval,
+                        success=_cs2_exec_success,
+                        error_message=_cs2_exec_error,
+                    )
+                except Exception as _cs2_finalize_err:
+                    logger.warning(
+                        "cs2_auto_execute_finalize_failed",
+                        approval_id=str(approval.id),
+                        incident_id=incident_id,
+                        error=str(_cs2_finalize_err),
+                    )
+
             _is_heartbeat = is_heartbeat_alertname(alertname)
             if can_auto_repair and not _is_heartbeat:
                 await _try_auto_repair_background(
@@ -1731,7 +1919,12 @@ async def _process_new_alert_background(
             record_alert_chain_success("alertmanager")
             return
 
-        analysis_result, ai_provider, raw_response, signoz_metrics, signoz_trace_url, ai_tokens, ai_cost = await openclaw.analyze_alert(alert_context)
+        analysis_result, ai_provider, raw_response, signoz_metrics, signoz_trace_url, ai_tokens, ai_cost = await _analyze_alertmanager_with_timeout(
+            openclaw,
+            alert_context,
+            alert_id=alert_id,
+            alertname=alertname,
+        )
 
         if analysis_result:
             risk_mapping = {
@@ -1841,8 +2034,15 @@ async def _process_new_alert_background(
                 and "NO_ACTION" not in (analysis_result.action_title or "")
                 and is_safe_kubectl_action(_cs3_kubectl)
             )
+            _cs3_auto_approval = None
+            _cs3_executor = None
+            _cs3_exec_success: bool | None = None
+            _cs3_exec_error: str | None = None
             if _cs3_can_auto:
                 try:
+                    from src.models.approval import ApprovalRequest, ApprovalStatus
+                    from src.services.approval_execution import ApprovalExecutionService
+
                     _cs3_auto_approval = ApprovalRequest(
                         action=approval_create.action,
                         description=approval_create.description,
@@ -1859,8 +2059,17 @@ async def _process_new_alert_background(
                             else "cs3_auto_confident_execution",
                         },
                     )
+                    _cs3_auto_approval.id = approval.id
                     _cs3_executor = ApprovalExecutionService()
                     _cs3_exec_success = await _cs3_executor.execute_approved_action(_cs3_auto_approval)
+                    try:
+                        await service.update_execution_status(approval.id, _cs3_exec_success)
+                    except Exception as _cs3_upd_err:
+                        logger.warning(
+                            "cs3_auto_execute_status_update_failed",
+                            approval_id=str(approval.id),
+                            error=str(_cs3_upd_err),
+                        )
                     logger.info(
                         "cs3_llm_auto_executed",
                         approval_id=str(approval.id),
@@ -1876,6 +2085,8 @@ async def _process_new_alert_background(
                         ),
                     )
                 except Exception as _cs3_exec_err:
+                    _cs3_exec_success = False if _cs3_auto_approval is not None else None
+                    _cs3_exec_error = str(_cs3_exec_err)
                     logger.warning("cs3_llm_auto_execute_failed", error=str(_cs3_exec_err))
 
             incident_id = await create_incident_for_approval(
@@ -1887,7 +2098,7 @@ async def _process_new_alert_background(
                 message=message,
                 source="alertmanager",
                 alertname=alertname,
-                alert_labels=alert_labels,
+                alert_labels=traced_alert_labels,
                 notification_type=notification_type,
                 alert_category=alert_category,
             )
@@ -1903,6 +2114,41 @@ async def _process_new_alert_background(
                     error=str(_meta_err),
                 )
 
+            await record_alertmanager_event(
+                project_id="awoooi",
+                alert_id=alert_id,
+                alertname=alertname,
+                severity=severity,
+                namespace=namespace,
+                target_resource=target_resource,
+                fingerprint=fingerprint,
+                stage="incident_linked",
+                notification_type=notification_type,
+                alert_category=alert_category,
+                incident_id=incident_id,
+                approval_id=str(approval.id),
+                repeat_count=1,
+                labels=traced_alert_labels,
+                annotations=alert_context.get("annotations", {}),
+            )
+
+            if _cs3_auto_approval is not None and _cs3_exec_success is not None:
+                try:
+                    _cs3_auto_approval.incident_id = incident_id
+                    _cs3_executor = _cs3_executor or ApprovalExecutionService()
+                    await _cs3_executor.finalize_auto_approved_execution(
+                        _cs3_auto_approval,
+                        success=_cs3_exec_success,
+                        error_message=_cs3_exec_error,
+                    )
+                except Exception as _cs3_finalize_err:
+                    logger.warning(
+                        "cs3_auto_execute_finalize_failed",
+                        approval_id=str(approval.id),
+                        incident_id=incident_id,
+                        error=str(_cs3_finalize_err),
+                    )
+
             root_cause = analysis_result.description or message
             estimated_downtime = blast.estimated_downtime if blast else "~30s"
             primary_responsibility = analysis_result.primary_responsibility or "COLLAB"
@@ -1978,11 +2224,17 @@ async def _process_new_alert_background(
         else:
             # LLM 失敗 - 使用預設值
             # 2026-04-27 Claude Sonnet 4.6: shadow-run Step1 — 補 metadata kwarg，讓 extra_metadata 可觀測
+            _matched_playbook_id_cs4 = await resolve_playbook_id_for_alert(
+                rule_id=str(rule_response.get("rule_id", "")),
+                alertname=alertname,
+                affected_services=[target_resource] if target_resource else [],
+                severity="medium",
+            )
             _approval_metadata_cs4 = {
                 "source": "fallback",
                 "confidence_score": None,
                 "is_rule_based": False,
-                "playbook_id": None,
+                "playbook_id": _matched_playbook_id_cs4,
             }
             fallback_create = ApprovalRequestCreate(
                 action="OBSERVE",
@@ -1997,6 +2249,7 @@ async def _process_new_alert_background(
                 dry_run_checks=[],
                 requested_by="OpenClaw (fallback)",
                 metadata=_approval_metadata_cs4,
+                matched_playbook_id=_matched_playbook_id_cs4,
             )
 
             approval = await service.create_approval_with_fingerprint(
@@ -2034,7 +2287,7 @@ async def _process_new_alert_background(
                 message=message,
                 source="alertmanager",
                 alertname=alertname,
-                alert_labels=alert_labels,
+                alert_labels=traced_alert_labels,
                 notification_type=notification_type,
                 alert_category=alert_category,
             )
@@ -2050,6 +2303,55 @@ async def _process_new_alert_background(
                     error=str(_meta_err),
                 )
 
+            await record_alertmanager_event(
+                project_id="awoooi",
+                alert_id=alert_id,
+                alertname=alertname,
+                severity=severity,
+                namespace=namespace,
+                target_resource=target_resource,
+                fingerprint=fingerprint,
+                stage="incident_linked",
+                notification_type=notification_type,
+                alert_category=alert_category,
+                incident_id=fallback_incident_id,
+                approval_id=str(approval.id),
+                repeat_count=1,
+                labels=traced_alert_labels,
+                annotations=alert_context.get("annotations", {}),
+            )
+
+            _is_heartbeat = is_heartbeat_alertname(alertname)
+            if can_auto_repair and not _is_heartbeat:
+                await _try_auto_repair_background(
+                    incident_id=fallback_incident_id,
+                    approval_id=str(approval.id),
+                    alert_type=alert_type,
+                    target_resource=target_resource,
+                    namespace=namespace,
+                )
+            elif not can_auto_repair and not _is_heartbeat:
+                from src.repositories.alert_operation_log_repository import get_alert_operation_log_repository
+                _op_log_fallback = get_alert_operation_log_repository()
+                await _op_log_fallback.append(
+                    "GUARDRAIL_BLOCKED",
+                    incident_id=fallback_incident_id,
+                    approval_id=str(approval.id),
+                    actor="prometheus-rule",
+                    action_detail=f"Prometheus rule 設定 auto_repair=false，fallback 轉人工: {alertname}",
+                    success=False,
+                    context={"alertname": alertname, "auto_repair_flag": False},
+                )
+                await _escalate_auto_repair_unavailable(
+                    incident_id=fallback_incident_id,
+                    approval_id=str(approval.id),
+                    alert_type=alert_type,
+                    target_resource=target_resource,
+                    namespace=namespace,
+                    failure_reason="Prometheus rule auto_repair=false，fallback 未進入自動修復評估",
+                    attempted_actions="llm_fallback -> guardrail:auto_repair_false -> emergency_intervention",
+                )
+
             await _push_to_telegram_background(
                 approval_id=str(approval.id),
                 risk_level="medium",
@@ -2139,6 +2441,7 @@ async def alertmanager_webhook(
     # (2026-04-08 Claude Sonnet 4.6 Asia/Taipei，ADR-062 Q9)
     # ==========================================================================
     _alert_labels = alert.labels or {}
+    _alert_annotations = alert.annotations or {}
     _alertname_for_log = _alert_labels.get("alertname", "UnknownAlert")
     # Q9: auto_repair flag — Rule=false 強制 HITL（不觸發自動修復背景任務）
     _can_auto_repair_by_rule = _alert_labels.get("auto_repair", "true").lower() == "true"
@@ -2154,6 +2457,7 @@ async def alertmanager_webhook(
                 "alert_id": alert_id,
                 "alertname": _alertname_for_log,
                 "labels": _alert_labels,
+                "annotations": _alert_annotations,
                 "auto_repair_flag": _can_auto_repair_by_rule,
             },
         )
@@ -2182,11 +2486,12 @@ async def alertmanager_webhook(
             telegram = get_telegram_gateway()
             # 解析 CI/CD 狀態
             stage = alert.labels.get("stage", "")
-            job_status = "success" if alert.labels.get("severity") == "info" else "running"
+            job_status = _cicd_job_status_from_alert(alert)
             commit_sha = alert.labels.get("commit", "")
             triggered_by = alert.labels.get("triggered_by", "CI")
             workflow_url = alert.annotations.get("workflow_url", "")
             summary = alert.annotations.get("summary", alertname)
+            detail_message = alert.annotations.get("description", "")
 
             await telegram.send_cicd_progress(
                 job_name=summary,
@@ -2194,6 +2499,8 @@ async def alertmanager_webhook(
                 stage=stage,
                 commit_sha=commit_sha,
                 triggered_by=triggered_by,
+                duration_seconds=_cicd_duration_seconds_from_alert(alert),
+                message=detail_message,
                 workflow_url=workflow_url,
             )
 
@@ -2293,6 +2600,22 @@ async def alertmanager_webhook(
         target=target_resource,
         fingerprint=fingerprint,
     )
+    background_tasks.add_task(
+        record_alertmanager_event,
+        project_id="awoooi",
+        alert_id=alert_id,
+        alertname=alertname,
+        severity=severity,
+        namespace=namespace,
+        target_resource=target_resource,
+        fingerprint=fingerprint,
+        stage="received",
+        notification_type=notification_type,
+        alert_category=alert_category,
+        source_url=alert.generatorURL,
+        labels=dict(alert.labels) if alert.labels else {},
+        annotations=dict(alert.annotations) if alert.annotations else {},
+    )
 
     # ==========================================================================
     # ADR-076: 告警聚合引擎 — 5 分鐘滑動視窗，防止告警風暴
@@ -2375,6 +2698,26 @@ async def alertmanager_webhook(
                     hit_count=updated_approval.hit_count,
                     reason="Converged alert - Telegram already sent for this fingerprint",
                 )
+                background_tasks.add_task(
+                    record_alertmanager_event,
+                    project_id="awoooi",
+                    alert_id=alert_id,
+                    alertname=alertname,
+                    severity=severity,
+                    namespace=namespace,
+                    target_resource=target_resource,
+                    fingerprint=fingerprint,
+                    stage="converged",
+                    notification_type=notification_type,
+                    alert_category=alert_category,
+                    incident_id=getattr(updated_approval, "incident_id", None),
+                    approval_id=str(updated_approval.id),
+                    repeat_count=updated_approval.hit_count,
+                    is_duplicate=True,
+                    source_url=alert.generatorURL,
+                    labels=dict(alert.labels) if alert.labels else {},
+                    annotations=dict(alert.annotations) if alert.annotations else {},
+                )
 
                 return AlertResponse(
                     success=True,
@@ -2402,10 +2745,27 @@ async def alertmanager_webhook(
                 message=message,
                 source="alertmanager",
                 alertname=alertname,
-                alert_labels=alert.labels,
+                alert_labels={**alert.labels, "fingerprint": fingerprint, "alert_id": alert_id},
                 notification_type="TYPE-1",
                 alert_category=alert_category,
             )
+            background_tasks.add_task(
+                record_alertmanager_event,
+                project_id="awoooi",
+                alert_id=alert_id,
+                alertname=alertname,
+                severity=severity,
+                namespace=namespace,
+                target_resource=target_resource,
+                fingerprint=fingerprint,
+                stage="incident_linked",
+                notification_type="TYPE-1",
+                alert_category=alert_category,
+                incident_id=_info_incident_id,
+                source_url=alert.generatorURL,
+                labels={**alert.labels, "fingerprint": fingerprint, "alert_id": alert_id},
+                annotations=dict(alert.annotations) if alert.annotations else {},
+            )
             # 2026-04-15 ogt: TYPE-1 純資訊告警建立後立即關閉
             # 設計原則: backup/heartbeat/info 告警無需追蹤狀態，通知即完成
             # 防止 incidents 表無限累積 INVESTIGATING 記錄（ADR-073 漏洞修補）
@@ -2425,7 +2785,7 @@ async def alertmanager_webhook(
             record_alert_chain_success("alertmanager")
             return AlertResponse(
                 success=True,
-                message=f"✅ TYPE-1 純資訊告警已通知 (no LLM)",
+                message="✅ TYPE-1 純資訊告警已通知 (no LLM)",
                 alert_id=alert_id,
                 approval_created=False,
             )
@@ -2437,6 +2797,23 @@ async def alertmanager_webhook(
                 fingerprint=fingerprint,
                 ttl_seconds=ALERTMANAGER_LLM_INFLIGHT_LOCK_TTL_SECONDS,
             )
+            background_tasks.add_task(
+                record_alertmanager_event,
+                project_id="awoooi",
+                alert_id=alert_id,
+                alertname=alertname,
+                severity=severity,
+                namespace=namespace,
+                target_resource=target_resource,
+                fingerprint=fingerprint,
+                stage="llm_inflight_suppressed",
+                notification_type=notification_type,
+                alert_category=alert_category,
+                is_duplicate=True,
+                source_url=alert.generatorURL,
+                labels=dict(alert.labels) if alert.labels else {},
+                annotations=dict(alert.annotations) if alert.annotations else {},
+            )
             return AlertResponse(
                 success=True,
                 message="🛡️ 告警已由同指紋背景 AI 分析處理中，跳過重複 LLM 呼叫",

apps/api/src/core/config.py

@@ -609,6 +609,70 @@ class Settings(BaseSettings):
             "(X-AwoooP-Operator-Key header)"
         ),
     )
+    ENABLE_AWOOOP_ANSIBLE_CHECK_MODE_WORKER: bool = Field(
+        default=False,
+        description=(
+            "True=consume ansible_candidate_matched AOL rows and run "
+            "ansible-playbook --check --diff only. Apply remains disabled."
+        ),
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_INTERVAL_SECONDS: int = Field(
+        default=300,
+        ge=60,
+        description="AwoooP Ansible check-mode worker polling interval.",
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_BATCH_LIMIT: int = Field(
+        default=1,
+        ge=1,
+        le=5,
+        description="Maximum Ansible check-mode candidates claimed per worker tick.",
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_TIMEOUT_SECONDS: int = Field(
+        default=180,
+        ge=30,
+        le=600,
+        description="Timeout for one ansible-playbook --check --diff execution.",
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_STARTUP_SLEEP_SECONDS: int = Field(
+        default=120,
+        ge=0,
+        le=900,
+        description="Delay before the check-mode worker first tick after API startup.",
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_TRANSPORT_PROFILE: str = Field(
+        default="ssh_mcp",
+        description=(
+            "SSH transport profile used by Ansible check-mode. Production uses "
+            "the existing ssh-mcp key so repair-bot forced-command remains reserved "
+            "for whitelist repairs."
+        ),
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_SSH_KEY_PATH: str = Field(
+        default="/run/secrets/ssh_mcp_key",
+        description="Private key path for Ansible check-mode SSH transport.",
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_KNOWN_HOSTS_PATH: str = Field(
+        default="/etc/ssh-mcp/known_hosts",
+        description="known_hosts path for Ansible check-mode SSH transport.",
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_CANDIDATE_MAX_AGE_HOURS: int = Field(
+        default=24,
+        ge=1,
+        le=168,
+        description=(
+            "Only recent Ansible candidate audit rows are eligible for automatic "
+            "check-mode claims; older backlog remains visible but is not drained as noise."
+        ),
+    )
+    AWOOOP_ANSIBLE_CHECK_MODE_TRANSPORT_COOLDOWN_SECONDS: int = Field(
+        default=21_600,
+        ge=300,
+        le=86_400,
+        description=(
+            "Cooldown after transport-level check-mode blockers such as "
+            "forced-command repair SSH denial."
+        ),
+    )
 
     # ==========================================================================
     # 統帥鐵律：禁止 SQLite (AWOOOI 憲法)

apps/api/src/core/constants.py

@@ -37,8 +37,8 @@ REDIS_KEY_DECISION = "decision:"
 APPROVAL_TO_INCIDENT_STATUS = {
     "pending": "investigating",
     "approved": "resolved",
-    "rejected": "rejected",
-    "expired": "expired",
+    "rejected": "escalated",
+    "expired": "escalated",
 }
 
 # Incident 狀態 → 是否活躍

apps/api/src/core/unit_of_work.py

@@ -17,6 +17,7 @@ PostgreSQL 事務管理器，確保多表操作原子性。
 from typing import Any
 
 import structlog
+from sqlalchemy import text
 from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
 
 logger = structlog.get_logger(__name__)
@@ -49,14 +50,20 @@ class UnitOfWork:
         - Redis 操作失敗時必須手動呼叫 rollback()
     """
 
-    def __init__(self, session_factory: async_sessionmaker[AsyncSession]):
+    def __init__(
+        self,
+        session_factory: async_sessionmaker[AsyncSession],
+        project_id: str | None = None,
+    ):
         """
         初始化 UnitOfWork
 
         Args:
             session_factory: SQLAlchemy async session factory
+            project_id: RLS project context. None means contextvar/default awoooi.
         """
         self._session_factory = session_factory
+        self._project_id = project_id
         self._session: AsyncSession | None = None
         self._committed = False
 
@@ -74,9 +81,18 @@ class UnitOfWork:
 
     async def __aenter__(self) -> "UnitOfWork":
         """進入事務"""
+        from src.core.context import get_current_project_id
+
         self._session = self._session_factory()
+        effective_pid = (
+            self._project_id if self._project_id is not None else get_current_project_id()
+        )
+        await self._session.execute(
+            text("SELECT set_config('app.project_id', :pid, TRUE)"),
+            {"pid": effective_pid},
+        )
         self._committed = False
-        logger.debug("uow_started")
+        logger.debug("uow_started", project_id=effective_pid)
         return self
 
     async def __aexit__(

apps/api/src/db/awooop_models.py

@@ -635,6 +635,13 @@ class AwoooPConversationEvent(Base):
     content_type: Mapped[str] = mapped_column(String(32), nullable=False, default="text")
     content_hash: Mapped[str | None] = mapped_column(String(64), nullable=True)
     content_preview: Mapped[str | None] = mapped_column(String(256), nullable=True)
+    content_redacted: Mapped[str | None] = mapped_column(Text, nullable=True)
+    redaction_version: Mapped[str] = mapped_column(
+        String(32), nullable=False, server_default=text("'audit_sink_v1'")
+    )
+    source_envelope: Mapped[dict[str, Any]] = mapped_column(
+        JSONB, nullable=False, server_default=text("'{}'::jsonb")
+    )
     attachment_sha256: Mapped[str | None] = mapped_column(String(64), nullable=True)
     is_duplicate: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
     provider_ts: Mapped[datetime | None] = mapped_column(nullable=True)
@@ -680,6 +687,13 @@ class AwoooPOutboundMessage(Base):
     message_type: Mapped[str] = mapped_column(String(32), nullable=False)
     content_hash: Mapped[str | None] = mapped_column(String(64), nullable=True)
     content_preview: Mapped[str | None] = mapped_column(String(256), nullable=True)
+    content_redacted: Mapped[str | None] = mapped_column(Text, nullable=True)
+    redaction_version: Mapped[str] = mapped_column(
+        String(32), nullable=False, server_default=text("'audit_sink_v1'")
+    )
+    source_envelope: Mapped[dict[str, Any]] = mapped_column(
+        JSONB, nullable=False, server_default=text("'{}'::jsonb")
+    )
     provider_message_id: Mapped[str | None] = mapped_column(String(64), nullable=True)
     send_status: Mapped[str] = mapped_column(String(16), nullable=False, default="pending")
     send_error: Mapped[str | None] = mapped_column(Text, nullable=True)

apps/api/src/db/base.py

@@ -106,10 +106,13 @@ async def get_db() -> AsyncGenerator[AsyncSession, None]:
     factory = get_session_factory()
     async with factory() as session:
         try:
+            from src.core.context import get_current_project_id
+
             # AwoooP Phase 2.3 (2026-05-04 ogt): SET LOCAL app.project_id 讓 RLS Policy 生效
-            # 預設 'awoooi'，多租戶路由將在 middleware 注入實際 project_id
+            # 預設 'awoooi'，多租戶路由將透過 contextvar 注入實際 project_id
             await session.execute(
-                text("SELECT set_config('app.project_id', 'awoooi', TRUE)")
+                text("SELECT set_config('app.project_id', :pid, TRUE)"),
+                {"pid": get_current_project_id()},
             )
             yield session
             await session.commit()
@@ -154,6 +157,9 @@ async def get_db_context(project_id: str | None = None) -> AsyncGenerator[AsyncS
 # Initialization
 # =============================================================================
 
+_DB_BOOTSTRAP_LOCK_NAME = "awoooi:init_db:ddl"
+
+
 async def init_db() -> None:
     """
     Initialize database tables
@@ -162,6 +168,28 @@ async def init_db() -> None:
     """
     engine = get_engine()
 
+    async with engine.connect() as lock_conn:
+        # 2026-05-24 ogt + Codex: 兩個 API replica 同時啟動時，PostgreSQL 會在
+        # ALTER TABLE ... IF NOT EXISTS 上互相等待並 deadlock。整段 bootstrap
+        # DDL 必須序列化，避免 rollout 因一個 pod CrashLoop 變成 1/2 ready。
+        await lock_conn.execute(
+            text("SELECT pg_advisory_lock(hashtext(:lock_name))"),
+            {"lock_name": _DB_BOOTSTRAP_LOCK_NAME},
+        )
+        try:
+            await _run_init_db_ddl(engine)
+        finally:
+            await lock_conn.execute(
+                text("SELECT pg_advisory_unlock(hashtext(:lock_name))"),
+                {"lock_name": _DB_BOOTSTRAP_LOCK_NAME},
+            )
+
+
+async def _run_init_db_ddl(engine: AsyncEngine) -> None:
+    """
+    Run idempotent DB bootstrap DDL while caller holds the bootstrap advisory lock.
+    """
+
     # 2026-04-15 ogt: 多 replica 並行啟動競爭修復
     # 問題：單一大 transaction 裡兩個 pod 同時建 table → 其中一個 CREATE INDEX 失敗
     #       PostgreSQL 中 transaction 內任何錯誤導致整個 transaction ROLLBACK

apps/api/src/db/models.py

@@ -633,6 +633,8 @@ class AlertOperationLog(Base):
             "RESOLVED", "SILENCED", "ESCALATED", "GUARDRAIL_BLOCKED",
             "PRE_FLIGHT_PASSED", "PRE_FLIGHT_FAILED", "BACKUP_TRIGGERED",
             "BACKUP_COMPLETED", "BACKUP_FAILED", "APPROVAL_ESCALATED", "CHANGE_APPLIED",
+            "NOTIFICATION_CLASSIFIED", "MANUAL_FIX_RECORDED", "KM_CONVERTED",
+            "PLAYBOOK_DRAFT_CREATED", "STATE_GUARD_BLOCKED",
             name="alert_event_type", create_type=False,
         ),
         nullable=False, index=True,

apps/api/src/hermes/nl_gateway.py

@@ -23,6 +23,7 @@ from src.db.base import get_db_context
 from src.hermes.agent_loader import get_agent_system_prompt
 from src.hermes.display_names import DEFAULT_AGENT, format_response_header
 from src.hermes.safety_hooks import is_dangerous_input, is_mutate_intent
+from src.services.ollama_endpoint_resolver import resolve_ollama_order
 
 logger = structlog.get_logger(__name__)
 
@@ -261,44 +262,48 @@ async def process_nl_message(
 
     t0 = time.monotonic()
 
-    # 呼叫 Ollama 本地模型（111，零費用，按 agent 選模型）
+    # 呼叫 Ollama 模型（GCP-A → GCP-B → 111，零費用，按 agent 選模型）
     model = _pick_model(agent_name)
     success = False
     error_type: str | None = None
-    try:
-        from src.services.ollama_endpoint_resolver import resolve_ollama_endpoint
-
-        ollama_base = resolve_ollama_endpoint("hermes")
-        async with httpx.AsyncClient(timeout=_OLLAMA_TIMEOUT) as _hc:
-            resp = await _hc.post(
-                f"{ollama_base}/api/chat",
-                json={
-                    "model": model,
-                    "messages": [
-                        {"role": "system", "content": system_prompt},
-                        {"role": "user",   "content": prompt_with_ctx},
-                    ],
-                    "stream": False,
-                    "options": {"num_predict": 1500, "temperature": 0.3},
-                },
-            )
-            resp.raise_for_status()
-            result_text = resp.json().get("message", {}).get("content", "")
-
-        result_text = _strip_think_tags(result_text)
-        if not result_text:
-            result_text = "_Agent 回應為空，請稍後再試。_"
-        success = True
-
-    except Exception as exc:
-        error_type = type(exc).__name__
-        logger.error(
-            "hermes_nl_ollama_error",
-            error=str(exc),
-            agent=agent_name,
-            model=model,
-            exc_type=error_type,
-        )
+    result_text = ""
+    async with httpx.AsyncClient(timeout=_OLLAMA_TIMEOUT) as _hc:
+        for endpoint in resolve_ollama_order("hermes"):
+            if not endpoint.url:
+                continue
+            try:
+                resp = await _hc.post(
+                    f"{endpoint.url}/api/chat",
+                    json={
+                        "model": model,
+                        # Keep Hermes responses in message.content across Ollama 0.24+.
+                        "think": False,
+                        "messages": [
+                            {"role": "system", "content": system_prompt},
+                            {"role": "user",   "content": prompt_with_ctx},
+                        ],
+                        "stream": False,
+                        "options": {"num_predict": 1500, "temperature": 0.3},
+                    },
+                )
+                resp.raise_for_status()
+                result_text = resp.json().get("message", {}).get("content", "")
+                result_text = _strip_think_tags(result_text)
+                if not result_text:
+                    result_text = "_Agent 回應為空，請稍後再試。_"
+                success = True
+                break
+            except Exception as exc:
+                error_type = type(exc).__name__
+                logger.error(
+                    "hermes_nl_ollama_error",
+                    error=str(exc),
+                    agent=agent_name,
+                    model=model,
+                    provider=endpoint.provider_name,
+                    exc_type=error_type,
+                )
+    if not success:
         result_text = f"_Hermes 暫時無法連線（{error_type}），請稍後再試。_"
 
     latency_ms = int((time.monotonic() - t0) * 1000)

apps/api/src/jobs/awooop_ansible_check_mode_job.py

@@ -0,0 +1,44 @@
+"""AwoooP Ansible check-mode worker loop.
+
+Runs only when explicitly enabled by settings. The worker consumes pending
+``ansible_candidate_matched`` rows and records check-mode evidence; it never
+executes Ansible apply.
+"""
+
+from __future__ import annotations
+
+import asyncio
+
+import structlog
+
+from src.core.config import settings
+from src.services.awooop_ansible_check_mode_service import run_pending_check_modes_once
+
+logger = structlog.get_logger(__name__)
+
+
+async def run_awooop_ansible_check_mode_loop() -> None:
+    if not settings.ENABLE_AWOOOP_ANSIBLE_CHECK_MODE_WORKER:
+        logger.info("awooop_ansible_check_mode_worker_disabled")
+        return
+
+    logger.info(
+        "awooop_ansible_check_mode_worker_started",
+        interval_seconds=settings.AWOOOP_ANSIBLE_CHECK_MODE_INTERVAL_SECONDS,
+        batch_limit=settings.AWOOOP_ANSIBLE_CHECK_MODE_BATCH_LIMIT,
+        timeout_seconds=settings.AWOOOP_ANSIBLE_CHECK_MODE_TIMEOUT_SECONDS,
+    )
+    await asyncio.sleep(settings.AWOOOP_ANSIBLE_CHECK_MODE_STARTUP_SLEEP_SECONDS)
+
+    while True:
+        try:
+            result = await run_pending_check_modes_once(
+                limit=settings.AWOOOP_ANSIBLE_CHECK_MODE_BATCH_LIMIT,
+                timeout_seconds=settings.AWOOOP_ANSIBLE_CHECK_MODE_TIMEOUT_SECONDS,
+            )
+            if result.get("claimed") or result.get("blockers"):
+                logger.info("awooop_ansible_check_mode_worker_tick", **result)
+        except Exception as exc:
+            logger.warning("awooop_ansible_check_mode_worker_failed", error=str(exc))
+
+        await asyncio.sleep(settings.AWOOOP_ANSIBLE_CHECK_MODE_INTERVAL_SECONDS)

apps/api/src/jobs/hermes_kb_growth_worker.py

@@ -0,0 +1,308 @@
+"""
+Hermes KB Growth Worker
+=======================
+
+消費 governance_remediation_dispatch 中的 hermes_kb_growth_healthcheck work item，
+把 knowledge_degradation 告警推進成可審核的 KM 草稿。
+
+邊界：
+- 可以建立 REVIEW 狀態的 auto_runbook 草稿，讓 owner 在前端審核。
+- 不可以直接把 KM 標成 APPROVED / PUBLISHED。
+- 不修改 immutable ai_governance_events；流程進度寫回 dispatch.decision_context。
+
+2026-05-19 ogt + Codex: T90 Hermes KB growth healthcheck worker。
+"""
+
+from __future__ import annotations
+
+import asyncio
+from copy import deepcopy
+from typing import Any
+
+import structlog
+
+from src.db.base import get_db_context
+from src.db.models import GovernanceRemediationDispatch
+from src.models.knowledge import (
+    EntrySource,
+    EntryStatus,
+    EntryType,
+    KnowledgeEntry,
+    KnowledgeEntryCreate,
+)
+from src.repositories.governance_remediation_dispatch_repo import (
+    InvalidStatusTransition,
+    list_pending_by_executor,
+    transition_status,
+    update_decision_context,
+)
+from src.repositories.knowledge_repository import KnowledgeDBRepository
+
+logger = structlog.get_logger(__name__)
+
+EXECUTOR_TYPE = "hermes_kb_growth_healthcheck"
+DEFAULT_INTERVAL_SECONDS = 300
+DEFAULT_LIMIT = 20
+
+
+async def run_hermes_kb_growth_once(limit: int = DEFAULT_LIMIT) -> dict[str, int]:
+    """執行一輪 Hermes KB growth healthcheck。
+
+    Returns:
+        統計資訊，供 log / smoke test 判讀。
+    """
+    rows = await list_pending_by_executor(EXECUTOR_TYPE, limit=limit)
+    result = {
+        "scanned": len(rows),
+        "processed": 0,
+        "skipped": 0,
+        "failed": 0,
+    }
+
+    for row in rows:
+        try:
+            await _process_dispatch(row)
+            result["processed"] += 1
+        except InvalidStatusTransition as exc:
+            result["skipped"] += 1
+            logger.info(
+                "hermes_kb_growth_dispatch_skipped",
+                dispatch_id=row.id,
+                event_id=row.governance_event_id,
+                reason=str(exc),
+            )
+        except Exception as exc:
+            result["failed"] += 1
+            logger.exception(
+                "hermes_kb_growth_dispatch_failed",
+                dispatch_id=row.id,
+                event_id=row.governance_event_id,
+                error=str(exc),
+            )
+            await _mark_failed_if_started(row.id, str(exc))
+
+    if any(result.values()):
+        logger.info("hermes_kb_growth_once_completed", **result)
+    return result
+
+
+async def run_hermes_kb_growth_loop(
+    interval_seconds: int = DEFAULT_INTERVAL_SECONDS,
+    limit: int = DEFAULT_LIMIT,
+) -> None:
+    """背景 loop：定期消費 Hermes KB growth dispatch。"""
+    logger.info(
+        "hermes_kb_growth_loop_started",
+        interval_seconds=interval_seconds,
+        limit=limit,
+    )
+    while True:
+        try:
+            await run_hermes_kb_growth_once(limit=limit)
+        except asyncio.CancelledError:
+            raise
+        except Exception as exc:
+            logger.exception("hermes_kb_growth_loop_error", error=str(exc))
+        await asyncio.sleep(interval_seconds)
+
+
+async def _process_dispatch(row: GovernanceRemediationDispatch) -> None:
+    """處理單筆 pending dispatch，最後停在 waiting_owner_review。"""
+    dispatched = await transition_status(row.id, "pending", "dispatched")
+    executing = await transition_status(dispatched.id, "dispatched", "executing")
+
+    km_entry = await _create_or_get_km_review_draft(executing)
+    updated_context = _build_review_context(
+        executing.decision_context or {},
+        dispatch_id=executing.id,
+        governance_event_id=executing.governance_event_id,
+        km_entry_id=km_entry.id,
+    )
+    await update_decision_context(executing.id, updated_context)
+    await transition_status(executing.id, "executing", "succeeded")
+
+    logger.info(
+        "hermes_kb_growth_review_draft_ready",
+        dispatch_id=executing.id,
+        event_id=executing.governance_event_id,
+        km_entry_id=km_entry.id,
+        workflow_stage="waiting_owner_review",
+    )
+
+
+async def _create_or_get_km_review_draft(
+    dispatch: GovernanceRemediationDispatch,
+) -> KnowledgeEntry:
+    """以 governance event tag 做冪等，建立或取得 REVIEW 狀態 KM 草稿。"""
+    dispatch_tag = f"dispatch:{dispatch.id}"
+    event_tag = f"governance_event:{dispatch.governance_event_id}"
+    payload = _build_km_review_entry_payload(dispatch)
+
+    async with get_db_context() as db:
+        repo = KnowledgeDBRepository(db)
+        existing, _ = await repo.list_entries(tags=[event_tag], limit=1)
+        if existing:
+            return existing[0]
+        existing, _ = await repo.list_entries(tags=[dispatch_tag], limit=1)
+        if existing:
+            return existing[0]
+        return await repo.create(payload)
+
+
+def _build_km_review_entry_payload(
+    dispatch: GovernanceRemediationDispatch,
+) -> KnowledgeEntryCreate:
+    """把 governance dispatch 轉成待審核的 KM 草稿 payload。"""
+    context = dispatch.decision_context or {}
+    workflow = context.get("workflow") if isinstance(context.get("workflow"), dict) else {}
+    impact = workflow.get("impact") if isinstance(workflow.get("impact"), dict) else {}
+    extra = context.get("extra") if isinstance(context.get("extra"), dict) else {}
+    ownership = context.get("ownership") if isinstance(context.get("ownership"), dict) else {}
+    if not ownership and isinstance(extra.get("ownership"), dict):
+        ownership = extra["ownership"]
+
+    stale_count = _pick_first(impact, extra, key="stale_count")
+    total_count = _pick_first(impact, extra, key="total_count")
+    stale_ratio = _pick_first(impact, context, key="stale_ratio")
+    threshold = _pick_first(impact, context, key="threshold")
+    stale_days = _pick_first(impact, extra, key="stale_days")
+    lead_agent = ownership.get("lead_agent") or "Hermes"
+    human_owner = ownership.get("human_owner") or "KM owner / SRE owner"
+
+    content = "\n".join([
+        "# KM 健康檢查草稿",
+        "",
+        "## 來源",
+        f"- governance_event_id: {dispatch.governance_event_id}",
+        f"- dispatch_id: {dispatch.id}",
+        f"- executor_type: {dispatch.executor_type}",
+        "",
+        "## 影響摘要",
+        f"- stale_count: {_format_unknown(stale_count)}",
+        f"- total_count: {_format_unknown(total_count)}",
+        f"- stale_ratio: {_format_ratio(stale_ratio)}",
+        f"- threshold: {_format_ratio(threshold)}",
+        f"- stale_days: {_format_unknown(stale_days)}",
+        "",
+        "## AI 已完成",
+        "- Hermes 已接手 knowledge_degradation dispatch。",
+        "- 已產生 KM 更新草稿與 owner review work item。",
+        "- 尚未把任何條目標成 approved / published。",
+        "",
+        "## Owner 審核重點",
+        "- 優先反查最近被 Incident、Sentry、SigNoz、PlayBook 引用的 KM。",
+        "- 確認草稿內容沒有把過期處置方式寫回正式知識庫。",
+        "- 審核通過後再進入 km_writeback_after_approval。",
+        "",
+        "## 安全邊界",
+        "- writes_km_without_approval=false",
+        f"- lead_agent={lead_agent}",
+        f"- human_owner={human_owner}",
+    ])
+
+    return KnowledgeEntryCreate(
+        title=f"KM healthcheck review draft - {dispatch.governance_event_id[:8]}",
+        content=content,
+        entry_type=EntryType.AUTO_RUNBOOK,
+        category="AI治理",
+        tags=[
+            "governance:knowledge_degradation",
+            "workflow:kb_growth_healthcheck",
+            "stage:waiting_owner_review",
+            "agent:Hermes",
+            "needs_owner_review",
+            f"dispatch:{dispatch.id}",
+            f"governance_event:{dispatch.governance_event_id}",
+        ],
+        source=EntrySource.AI_EXTRACTED,
+        status=EntryStatus.REVIEW,
+        path_type="hermes_kb_growth_healthcheck",
+        created_by="hermes_kb_growth_worker",
+    )
+
+
+def _build_review_context(
+    context: dict[str, Any],
+    *,
+    dispatch_id: str,
+    governance_event_id: str,
+    km_entry_id: str,
+) -> dict[str, Any]:
+    """更新 dispatch read model，讓 Work Items/Telegram 可見目前停在 owner review。"""
+    updated = deepcopy(context)
+    workflow = updated.setdefault("workflow", {})
+    if not isinstance(workflow, dict):
+        workflow = {}
+        updated["workflow"] = workflow
+
+    stages = workflow.setdefault("stage_by_dispatch_status", {})
+    if not isinstance(stages, dict):
+        stages = {}
+        workflow["stage_by_dispatch_status"] = stages
+    stages.update({
+        "executing": "draft_km_updates",
+        "succeeded": "waiting_owner_review",
+        "failed": "needs_manual_km_triage",
+    })
+
+    workflow["current_stage"] = "waiting_owner_review"
+    workflow["next_action"] = "owner_review_km_draft"
+    workflow["needs_human_review"] = True
+    workflow["writes_km_without_approval"] = False
+    workflow["kb_draft_entry_id"] = km_entry_id
+
+    updated["next_action"] = "owner_review_km_draft"
+    updated["decision_path"] = "draft_created_waiting_owner_review"
+    updated["proposed_action"] = "Hermes 已建立 KM 更新草稿，等待 owner 審核"
+    updated["worker_result"] = {
+        "worker": "Hermes",
+        "executor_type": EXECUTOR_TYPE,
+        "dispatch_id": dispatch_id,
+        "governance_event_id": governance_event_id,
+        "km_draft_entry_id": km_entry_id,
+        "stage": "waiting_owner_review",
+        "status": "draft_created",
+        "writes_km_without_approval": False,
+    }
+    return updated
+
+
+async def _mark_failed_if_started(dispatch_id: str, error: str) -> None:
+    """若 worker 已取得 dispatch，將它收斂到 failed，保留錯誤。"""
+    for from_status in ("executing", "dispatched"):
+        try:
+            await transition_status(
+                dispatch_id,
+                from_status,
+                "failed",
+                last_error=error[:500],
+            )
+            return
+        except InvalidStatusTransition:
+            continue
+        except Exception as exc:
+            logger.warning(
+                "hermes_kb_growth_mark_failed_failed",
+                dispatch_id=dispatch_id,
+                from_status=from_status,
+                error=str(exc),
+            )
+            return
+
+
+def _pick_first(*sources: dict[str, Any], key: str) -> Any:
+    for source in sources:
+        if key in source:
+            return source[key]
+    return None
+
+
+def _format_unknown(value: Any) -> str:
+    return "unknown" if value is None else str(value)
+
+
+def _format_ratio(value: Any) -> str:
+    try:
+        return f"{float(value) * 100:.1f}%"
+    except (TypeError, ValueError):
+        return "unknown"

apps/api/src/jobs/incident_lifecycle_reconciler.py

@@ -0,0 +1,289 @@
+"""
+Incident Lifecycle Reconciler
+=============================
+
+把已有強證據的舊 stuck incident 收斂回 RESOLVED。
+
+範圍刻意保守：
+- auto_repair_executions.success = true
+- approval_records.status = EXECUTION_SUCCESS
+- approval_records.status = EXPIRED
+
+不處理單純 APPROVED / NO_ACTION / manual_required，避免把仍需人工的事件
+誤當作自動修復完成。
+"""
+
+from __future__ import annotations
+
+import asyncio
+from dataclasses import dataclass
+
+import httpx
+import structlog
+from sqlalchemy import text
+
+from src.core.config import settings
+from src.db.base import get_db_context
+from src.utils.timezone import now_taipei
+
+logger = structlog.get_logger(__name__)
+
+BATCH_LIMIT = 100
+INTERVAL_SECONDS = 1800
+_PROMETHEUS_TIMEOUT_SECONDS = 5.0
+
+
+@dataclass(frozen=True)
+class LifecycleCandidate:
+    incident_id: str
+    resolution_type: str
+    reason: str
+    direct_db_only: bool = False
+
+
+async def run_incident_lifecycle_reconciler_loop() -> None:
+    """每 30 分鐘收斂一小批已有完成證據的 stuck incident。"""
+    while True:
+        try:
+            resolved, errors = await reconcile_stuck_incidents()
+            if resolved > 0 or errors > 0:
+                logger.info(
+                    "incident_lifecycle_reconciler_done",
+                    resolved=resolved,
+                    errors=errors,
+                    batch_limit=BATCH_LIMIT,
+                )
+        except Exception as exc:
+            logger.warning("incident_lifecycle_reconciler_loop_failed", error=str(exc))
+
+        await asyncio.sleep(INTERVAL_SECONDS)
+
+
+async def reconcile_stuck_incidents(limit: int = BATCH_LIMIT) -> tuple[int, int]:
+    """
+    找出已完成但仍卡在 INVESTIGATING 的 incident，透過 IncidentService 統一路徑結案。
+
+    Returns:
+        (resolved_count, error_count)
+    """
+    candidates = await _fetch_candidates(limit)
+    remaining = max(0, limit - len(candidates))
+    if remaining > 0:
+        active_alertnames = await _fetch_active_alertnames()
+        if active_alertnames is not None:
+            candidates.extend(
+                await _fetch_inactive_or_duplicate_alert_candidates(
+                    limit=remaining,
+                    active_alertnames=active_alertnames,
+                    exclude_incident_ids={c.incident_id for c in candidates},
+                )
+            )
+
+    if not candidates:
+        return 0, 0
+
+    from src.services.incident_service import get_incident_service
+
+    incident_service = get_incident_service()
+    resolved = 0
+    errors = 0
+
+    for candidate in candidates:
+        try:
+            if candidate.direct_db_only:
+                result = await _resolve_db_only(candidate.incident_id)
+            else:
+                result = await incident_service.resolve_incident(
+                    candidate.incident_id,
+                    resolution_type=candidate.resolution_type,
+                    emit_postmortem=False,
+                )
+            if not result:
+                continue
+            resolved += 1
+            logger.info(
+                "incident_lifecycle_reconciled",
+                incident_id=candidate.incident_id,
+                reason=candidate.reason,
+                resolution_type=candidate.resolution_type,
+                direct_db_only=candidate.direct_db_only,
+            )
+        except Exception as exc:
+            errors += 1
+            logger.warning(
+                "incident_lifecycle_reconcile_failed",
+                incident_id=candidate.incident_id,
+                reason=candidate.reason,
+                error=str(exc),
+            )
+
+    return resolved, errors
+
+
+async def _fetch_active_alertnames() -> set[str] | None:
+    """Read current firing alertnames from Prometheus. None means fail-closed."""
+    try:
+        async with httpx.AsyncClient(timeout=_PROMETHEUS_TIMEOUT_SECONDS) as client:
+            response = await client.get(
+                f"{settings.PROMETHEUS_URL.rstrip('/')}/api/v1/query",
+                params={"query": 'ALERTS{alertstate="firing"}'},
+            )
+            response.raise_for_status()
+            payload = response.json()
+    except Exception as exc:
+        logger.warning("incident_lifecycle_active_alerts_fetch_failed", error=str(exc))
+        return None
+
+    result = payload.get("data", {}).get("result", [])
+    active_alertnames = {
+        item.get("metric", {}).get("alertname")
+        for item in result
+        if item.get("metric", {}).get("alertname")
+    }
+    logger.info(
+        "incident_lifecycle_active_alerts_loaded",
+        active_alert_count=len(active_alertnames),
+    )
+    return active_alertnames
+
+
+async def _resolve_db_only(incident_id: str) -> bool:
+    from src.repositories.incident_repository import get_incident_repository
+
+    now = now_taipei()
+    return await get_incident_repository().update_status(
+        incident_id=incident_id,
+        status="resolved",
+        updated_at=now,
+        resolved_at=now,
+    )
+
+
+async def _fetch_candidates(limit: int) -> list[LifecycleCandidate]:
+    async with get_db_context() as db:
+        result = await db.execute(
+            text(
+                """
+                WITH stale AS (
+                    SELECT
+                        i.incident_id,
+                        i.created_at,
+                        EXISTS (
+                            SELECT 1
+                            FROM auto_repair_executions are
+                            WHERE are.incident_id = i.incident_id
+                              AND are.success IS TRUE
+                        ) AS has_success_auto_repair,
+                        EXISTS (
+                            SELECT 1
+                            FROM approval_records ar
+                            WHERE ar.incident_id = i.incident_id
+                              AND ar.status::text = 'EXECUTION_SUCCESS'
+                        ) AS has_execution_success,
+                        EXISTS (
+                            SELECT 1
+                            FROM approval_records ar
+                            WHERE ar.incident_id = i.incident_id
+                              AND ar.status::text = 'EXPIRED'
+                        ) AS has_expired_approval
+                    FROM incidents i
+                    WHERE i.status = 'INVESTIGATING'
+                      AND i.created_at <= now() - interval '24 hours'
+                )
+                SELECT
+                    incident_id,
+                    CASE
+                        WHEN has_success_auto_repair THEN 'auto_repair'
+                        WHEN has_execution_success THEN 'auto_repair'
+                        ELSE 'timeout'
+                    END AS resolution_type,
+                    CASE
+                        WHEN has_success_auto_repair THEN 'auto_repair_execution_success'
+                        WHEN has_execution_success THEN 'approval_execution_success'
+                        ELSE 'approval_expired'
+                    END AS reason
+                FROM stale
+                WHERE has_success_auto_repair
+                   OR has_execution_success
+                   OR has_expired_approval
+                ORDER BY created_at DESC
+                LIMIT :limit
+                """
+            ),
+            {
+                "limit": limit,
+            },
+        )
+        rows = result.mappings().all()
+
+    return [
+        LifecycleCandidate(
+            incident_id=str(row["incident_id"]),
+            resolution_type=str(row["resolution_type"]),
+            reason=str(row["reason"]),
+        )
+        for row in rows
+    ]
+
+
+async def _fetch_inactive_or_duplicate_alert_candidates(
+    *,
+    limit: int,
+    active_alertnames: set[str],
+    exclude_incident_ids: set[str],
+) -> list[LifecycleCandidate]:
+    """
+    收斂 Alertmanager 已不再 firing 的舊 incident，以及同一 active alertname 的舊重複案。
+
+    若 Prometheus/Alertmanager 讀不到 active alertnames，上層會 fail-closed 不呼叫本函式。
+    """
+    active_list = list(active_alertnames) or ["__no_active_alertnames__"]
+    exclude_list = list(exclude_incident_ids) or ["__no_excluded_incidents__"]
+    async with get_db_context() as db:
+        result = await db.execute(
+            text(
+                """
+                WITH ranked AS (
+                    SELECT
+                        i.incident_id,
+                        i.alertname,
+                        i.created_at,
+                        row_number() OVER (
+                            PARTITION BY i.alertname
+                            ORDER BY i.created_at DESC, i.incident_id DESC
+                        ) AS rn
+                    FROM incidents i
+                    WHERE i.status = 'INVESTIGATING'
+                      AND i.created_at <= now() - interval '24 hours'
+                      AND NOT (i.incident_id = ANY(:exclude_incident_ids))
+                )
+                SELECT
+                    incident_id,
+                    CASE
+                        WHEN alertname = ANY(:active_alertnames)
+                            THEN 'active_duplicate_stale'
+                        ELSE 'inactive_alert_stale'
+                    END AS reason
+                FROM ranked
+                WHERE NOT (alertname = ANY(:active_alertnames) AND rn = 1)
+                ORDER BY created_at ASC
+                LIMIT :limit
+                """
+            ),
+            {
+                "active_alertnames": active_list,
+                "exclude_incident_ids": exclude_list,
+                "limit": limit,
+            },
+        )
+        rows = result.mappings().all()
+
+    return [
+        LifecycleCandidate(
+            incident_id=str(row["incident_id"]),
+            resolution_type="timeout",
+            reason=str(row["reason"]),
+            direct_db_only=True,
+        )
+        for row in rows
+    ]

apps/api/src/jobs/kb_rot_cleaner.py

@@ -28,7 +28,7 @@ from datetime import timedelta
 import structlog
 from sqlalchemy import select, update
 
-from src.db.base import get_session_factory
+from src.db.base import get_db_context
 from src.db.models import AiGovernanceEvent, KnowledgeEntryRecord
 from src.utils.timezone import now_taipei
 
@@ -129,7 +129,7 @@ class KbRotCleaner:
         rot_reasons: dict[str, list[str]] = {}
         total = 0
 
-        async with get_session_factory()() as session:
+        async with get_db_context() as session:
             # 只掃 active 狀態（非 archived）
             q = await session.execute(
                 select(KnowledgeEntryRecord).where(
@@ -193,7 +193,7 @@ class KbRotCleaner:
         if not result.stale_ids:
             return
 
-        async with get_session_factory()() as session:
+        async with get_db_context() as session:
             # 逐條更新（避免 bulk update 覆蓋 tags JSONB）
             q = await session.execute(
                 select(KnowledgeEntryRecord).where(
@@ -220,7 +220,7 @@ class KbRotCleaner:
     async def _save_event(self, result: RotScanResult) -> None:
         """寫 kb_stale 事件到 ai_governance_events。"""
         try:
-            async with get_session_factory()() as session:
+            async with get_db_context() as session:
                 event = AiGovernanceEvent(
                     event_type="kb_stale",
                     details=result.to_dict(),

apps/api/src/jobs/knowledge_decay_job.py

@@ -33,7 +33,7 @@ from datetime import timedelta
 import structlog
 from sqlalchemy import and_, select, update
 
-from src.db.base import get_session_factory
+from src.db.base import get_db_context
 from src.db.models import KnowledgeEntryRecord
 from src.models.knowledge import EntryStatus
 from src.utils.timezone import now_taipei
@@ -112,8 +112,7 @@ class KnowledgeDecayJob:
         cutoff = now_taipei() - timedelta(days=DECAY_AGE_DAYS)
         decayable_statuses = [EntryStatus.DRAFT.value, EntryStatus.REVIEW.value]
 
-        session_factory = get_session_factory()
-        async with session_factory() as db:
+        async with get_db_context() as db:
             # 查：30 天未引用（view_count=0）且 updated_at < cutoff 的 draft/review 條目
             stmt = select(KnowledgeEntryRecord).where(
                 and_(

apps/api/src/jobs/offline_replay_service.py

@@ -29,7 +29,7 @@ from datetime import timedelta
 import structlog
 from sqlalchemy import and_, select
 
-from src.db.base import get_session_factory
+from src.db.base import get_db_context
 from src.db.models import AgentSession, AiGovernanceEvent, AutoRepairExecution, IncidentEvidence
 from src.utils.timezone import now_taipei
 
@@ -109,9 +109,7 @@ class OfflineReplayService:
 
     async def _run_replay(self) -> OfflineReplayReport:
         cutoff = now_taipei() - timedelta(days=REPLAY_LOOKBACK_DAYS)
-        session_factory = get_session_factory()
-
-        async with session_factory() as db:
+        async with get_db_context() as db:
             # 1. 取最近 N 個有 AgentSession(coordinator) 的 Incident
             stmt = (
                 select(AgentSession.incident_id)
@@ -137,7 +135,7 @@ class OfflineReplayService:
             )
 
         results: list[IncidentReplayResult] = []
-        async with session_factory() as db:
+        async with get_db_context() as db:
             for incident_id in incident_ids:
                 r = await self._replay_one(db, incident_id)
                 results.append(r)

apps/api/src/main.py

@@ -31,20 +31,26 @@ from fastapi.responses import JSONResponse, Response
 from prometheus_client import CONTENT_TYPE_LATEST, generate_latest
 from sentry_sdk.integrations.fastapi import FastApiIntegration
 from sentry_sdk.integrations.starlette import StarletteIntegration
+from uvicorn.middleware.proxy_headers import ProxyHeadersMiddleware
 
 from src.api.v1 import agents as agents_v1  # Phase 9.5: Agent Teams API
 from src.api.v1 import ai as ai_v1
-from src.api.v1 import aider_events as aider_events_v1  # aider-watch v2 ADR-091
-from src.api.v1 import ai_governance as ai_governance_v1  # 2026-05-02: /governance 頁面 3 endpoints
+from src.api.v1 import (
+    ai_governance as ai_governance_v1,  # 2026-05-02: /governance 頁面 3 endpoints
+)
 from src.api.v1 import ai_slo as ai_slo_v1  # Phase 6 ADR-087: AI SLO 自我治理
+from src.api.v1 import aider_events as aider_events_v1  # aider-watch v2 ADR-091
 from src.api.v1 import aiops_kpi as aiops_kpi_v1  # ADR-090 § Phase 7 KPI Dashboard
-from src.api.v1 import aiops_timeline as aiops_timeline_v1  # 2026-04-27 Wave8-X3 B4 timeline endpoint
-from src.api.v1 import approvals as approvals_v1
+from src.api.v1 import (
+    aiops_timeline as aiops_timeline_v1,  # 2026-04-27 Wave8-X3 B4 timeline endpoint
+)
 from src.api.v1 import alert_operation_logs as alert_operation_logs_v1
+from src.api.v1 import approvals as approvals_v1
 from src.api.v1 import audit_logs as audit_logs_v1
 from src.api.v1 import auto_repair as auto_repair_v1  # #8: 自動升級決策
 from src.api.v1 import csrf as csrf_v1  # Phase 20: CSRF Protection
 from src.api.v1 import dashboard as dashboard_v1
+from src.api.v1 import drift as drift_v1  # Phase 25 P2: Config Drift Detection
 from src.api.v1 import errors as errors_v1  # #40: Sentry 錯誤 BFF API
 from src.api.v1 import (
     gitea_webhook as gitea_webhook_v1,  # ADR-059: Gitea → OpenClaw (GitHub → Gitea 遷移)
@@ -56,19 +62,20 @@ from src.api.v1 import incidents as incidents_v1  # Phase 6.4: Decision Proposal
 from src.api.v1 import knowledge as knowledge_v1  # KB Phase 1: Knowledge Base
 from src.api.v1 import learning as learning_v1  # Phase D-G P0: Learning API
 from src.api.v1 import metrics as metrics_v1  # Phase 7: Gold Metrics (真實血脈)
+from src.api.v1 import monitoring as monitoring_v1  # 2026-04-03: 監控工具狀態
+from src.api.v1 import notifications as notifications_v1  # 2026-04-10: 通知頻道狀態
+from src.api.v1 import (
+    platform as platform_v1,  # AwoooP Phase 4: Platform Shell（Shadow Mode）
+)
 from src.api.v1 import playbooks as playbooks_v1  # #7: Playbook 萃取
 from src.api.v1 import proposals as proposals_v1  # Phase 6.4h: Proposals CRUD API
+from src.api.v1 import rag as rag_v1  # Phase 33 ADR-067: RAG 知識庫
 from src.api.v1 import (
     sentry_webhook as sentry_webhook_v1,  # Phase 10.2.1: Sentry → Telegram
 )
 from src.api.v1 import (
     signoz_webhook as signoz_webhook_v1,  # Phase 21: SignOz → Telegram (ADR-037)
 )
-from src.api.v1 import drift as drift_v1  # Phase 25 P2: Config Drift Detection
-from src.api.v1 import platform as platform_v1  # AwoooP Phase 4: Platform Shell（Shadow Mode）
-from src.api.v1 import rag as rag_v1  # Phase 33 ADR-067: RAG 知識庫
-from src.api.v1 import monitoring as monitoring_v1  # 2026-04-03: 監控工具狀態
-from src.api.v1 import notifications as notifications_v1  # 2026-04-10: 通知頻道狀態
 from src.api.v1 import stats as stats_v1  # Phase 6.5: Statistics Analytics
 from src.api.v1 import telegram as telegram_v1  # Phase 5.4: Telegram Gateway
 from src.api.v1 import telegram_webhook as telegram_webhook_v1  # ADR-094: Webhook入口
@@ -76,11 +83,13 @@ from src.api.v1 import terminal as terminal_v1  # Phase 19.1: Omni-Terminal SSE
 from src.api.v1 import timeline as timeline_v1
 from src.api.v1 import webhooks as webhooks_v1
 from src.core.config import settings
-from src.core.feature_flags import aiops_flags  # ADR-080: AI 自主化飛輪 feature flags 啟動驗證
 from src.core.http_client import close_all_http_clients, init_all_http_clients
 from src.core.logging import get_logger, setup_logging
-from src.core.redis_client import close_redis_pool, init_redis_pool
-from src.services.flywheel_stats_service import get_flywheel_stats_service
+from src.core.redis_client import (
+    close_redis_pool,
+    close_worker_redis_pool,
+    init_redis_pool,
+)
 from src.core.sse import get_publisher
 from src.core.telemetry import setup_telemetry, shutdown_telemetry
 
@@ -92,7 +101,10 @@ from src.routers import proposals as proposals_router
 
 # Legacy route imports (to be migrated)
 from src.routes import agent, notifications, pipelines, plugins
+from src.services.adr100_slo_metrics_service import get_adr100_slo_metrics_service
+from src.services.alert_chain_metrics_service import get_alert_chain_metrics_service
 from src.services.executor import close_executor
+from src.services.flywheel_stats_service import get_flywheel_stats_service
 
 # Phase 5: OpenClaw AI Engine
 from src.services.openclaw import close_openclaw
@@ -267,16 +279,21 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-05 ogt: 重開機後 Redis 清空，從 DB restore 未解決的 incidents
     # 統帥批准: 數據必須長久記錄，重開機後自動恢復 Working Memory
     try:
-        from src.services.incident_service import get_incident_service
+        from sqlalchemy import select
+
         from src.db.base import get_db_context
         from src.db.models import IncidentRecord
-        from sqlalchemy import select
+        from src.models.incident import IncidentStatus
+        from src.services.incident_service import get_incident_service
 
         incident_service = get_incident_service()
         async with get_db_context() as db:
             result = await db.execute(
                 select(IncidentRecord).where(
-                    IncidentRecord.status.in_(["investigating", "mitigating"])
+                    IncidentRecord.status.in_([
+                        IncidentStatus.INVESTIGATING,
+                        IncidentStatus.MITIGATING,
+                    ])
                 )
             )
             records = result.scalars().all()
@@ -284,31 +301,16 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         restored = 0
         for record in records:
             try:
-                from src.models.incident import Incident
-                incident = Incident(
-                    incident_id=record.incident_id,
-                    status=record.status,
-                    severity=record.severity,
-                    signals=record.signals or [],
-                    affected_services=record.affected_services or [],
-                    decision_chain=record.decision_chain,
-                    proposal_ids=record.proposal_ids or [],
-                    outcome=record.outcome,
-                    created_at=record.created_at,
-                    updated_at=record.updated_at,
-                    resolved_at=record.resolved_at,
-                    closed_at=record.closed_at,
-                    ttl_days=record.ttl_days,
-                    vectorized=record.vectorized,
-                    # ADR-073: 分類欄位必須還原，否則 KM 寫入時全為 "unknown"
-                    notification_type=record.notification_type,
-                    alert_category=record.alert_category,
-                )
+                incident = incident_service._record_to_incident(record)
                 if await incident_service.save_to_working_memory(incident):
                     restored += 1
-            except Exception:
+            except Exception as record_error:
                 # 舊資料 source 值不合法（node-exporter 等）→ 跳過
-                pass
+                logger.warning(
+                    "working_memory_warmup_record_skipped",
+                    incident_id=getattr(record, "incident_id", None),
+                    error=str(record_error),
+                )
 
         logger.info("working_memory_warmed_up", restored=restored, total=len(records))
     except Exception as e:
@@ -351,7 +353,9 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         logger.warning("playbook_pg_backfill_schedule_failed", error=str(e))
 
     try:
-        from src.services.playbook_embedding_service import ensure_playbook_embeddings_indexed
+        from src.services.playbook_embedding_service import (
+            ensure_playbook_embeddings_indexed,
+        )
         asyncio.create_task(ensure_playbook_embeddings_indexed())
         logger.info("playbook_embedding_indexing_scheduled")
     except Exception as e:
@@ -499,6 +503,40 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     except Exception as e:
         logger.warning("approval_timeout_resolver_schedule_failed", error=str(e))
 
+    # T73: 已有完成證據但仍卡在 INVESTIGATING 的舊 incident 小批次收斂。
+    # 僅處理 auto-repair success / approval EXECUTION_SUCCESS / approval EXPIRED，
+    # 不自動關閉 manual_required 或單純 APPROVED 事件。
+    try:
+        from src.jobs.incident_lifecycle_reconciler import (
+            INTERVAL_SECONDS as INCIDENT_LIFECYCLE_RECONCILER_INTERVAL,
+        )
+        from src.jobs.incident_lifecycle_reconciler import (
+            run_incident_lifecycle_reconciler_loop,
+        )
+        asyncio.create_task(run_incident_lifecycle_reconciler_loop())
+        logger.info(
+            "incident_lifecycle_reconciler_scheduled",
+            interval_sec=INCIDENT_LIFECYCLE_RECONCILER_INTERVAL,
+        )
+    except Exception as e:
+        logger.warning("incident_lifecycle_reconciler_schedule_failed", error=str(e))
+
+    # AwoooP Ansible check-mode worker.
+    # 只執行 ansible-playbook --check --diff 並回寫 automation_operation_log；
+    # apply 仍必須走 approval gate，本 worker 不寫 auto_repair_executions。
+    try:
+        from src.jobs.awooop_ansible_check_mode_job import (
+            run_awooop_ansible_check_mode_loop,
+        )
+        asyncio.create_task(run_awooop_ansible_check_mode_loop())
+        logger.info(
+            "awooop_ansible_check_mode_worker_scheduled",
+            enabled=settings.ENABLE_AWOOOP_ANSIBLE_CHECK_MODE_WORKER,
+            interval_seconds=settings.AWOOOP_ANSIBLE_CHECK_MODE_INTERVAL_SECONDS,
+        )
+    except Exception as e:
+        logger.warning("awooop_ansible_check_mode_worker_schedule_failed", error=str(e))
+
     # ADR-083 Phase 3: Evolver Agent（每日）— Playbook 自動合併 + 低信任封存
     # 2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 3 初始建立
     try:
@@ -510,7 +548,9 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
 
     # ADR-104 T2: LLM Playbook DRAFT governance（每小時）
     try:
-        from src.jobs.playbook_generation_governance_job import run_playbook_generation_governance_loop
+        from src.jobs.playbook_generation_governance_job import (
+            run_playbook_generation_governance_loop,
+        )
         asyncio.create_task(run_playbook_generation_governance_loop())
         logger.info(
             "playbook_generation_governance_loop_scheduled",
@@ -554,11 +594,11 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-27 P3.1-T3 by Claude
     try:
         from src.utils.timezone import now_taipei
-        from datetime import datetime as _dt
 
         async def _run_kb_rot_cleaner_loop() -> None:
-            from src.jobs.kb_rot_cleaner import get_kb_rot_cleaner
             import asyncio as _asyncio
+
+            from src.jobs.kb_rot_cleaner import get_kb_rot_cleaner
             while True:
                 try:
                     now = now_taipei()
@@ -649,14 +689,24 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     except Exception as e:
         logger.warning("governance_dispatcher_schedule_failed", error=str(e))
 
+    # T90 2026-05-19 ogt + Codex: Hermes KB growth worker（每 5 分鐘）
+    # 消費 knowledge_degradation 的 hermes_kb_growth_healthcheck dispatch，
+    # 只產生 REVIEW 草稿並停在 owner review，不直接批准或發布 KM。
+    try:
+        from src.jobs.hermes_kb_growth_worker import run_hermes_kb_growth_loop
+        asyncio.create_task(run_hermes_kb_growth_loop())
+        logger.info("hermes_kb_growth_worker_scheduled", interval_sec=300)
+    except Exception as e:
+        logger.warning("hermes_kb_growth_worker_schedule_failed", error=str(e))
+
     # 2026-04-25 P1.2 by Claude Engineer-A2 — failover 整合到 ai_router + lifespan
     # OllamaFailoverManager + OllamaAutoRecoveryService 飛輪接線：
     #   failover 切換時 → recovery_callback → set_current_primary → Redis 持久化
     #   recovery service 每 30s 檢查 → 111 連續 3 次 HEALTHY → 自動切回 → clear_cache
     # 順序：先取 singleton → wire callback → 啟動 recovery service（才能接收 callback）
     try:
-        from src.services.ollama_failover_manager import get_ollama_failover_manager
         from src.services.ollama_auto_recovery import get_ollama_auto_recovery_service
+        from src.services.ollama_failover_manager import get_ollama_failover_manager
 
         _failover_mgr = get_ollama_failover_manager()
         _recovery_svc = get_ollama_auto_recovery_service()
@@ -669,8 +719,8 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         # alerter 還沒注入 Redis → dedup fail-open，告警會送出且無 dedup 保護（重複告警風險）
         # 修法：configure_alerter() 提前到 start() 之前；Redis pool 在 lifespan 早期已就緒
         try:
-            from src.services.failover_alerter import configure_alerter
             from src.core.redis_client import get_redis
+            from src.services.failover_alerter import configure_alerter
             configure_alerter(get_redis())
             logger.info("failover_alerter_configured")
         except Exception as _alerter_err:
@@ -754,6 +804,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
 
     # Phase 6.1: 關閉 Signal Worker (先關閉 Consumer)
     await close_signal_worker()
+    await close_worker_redis_pool()
     await publisher.stop()
     await close_executor()
     await close_openclaw()
@@ -806,11 +857,8 @@ else:
 # Middleware
 # =============================================================================
 
-# 2026-04-03 ogt: Nginx 反向代理修正 — 讓 FastAPI 信任 X-Forwarded-Proto
-# 解決問題: /api/v1/knowledge (無結尾斜線) 307 redirect 產生 http:// Location
-# 原因: FastAPI 不知道自己在 HTTPS 後面，redirect 回 http://
-# 效果: 有了此中間件，307 Location 會是 https://
-from uvicorn.middleware.proxy_headers import ProxyHeadersMiddleware
+# 2026-04-03 ogt: Nginx 反向代理修正 — 讓 FastAPI 信任 X-Forwarded-Proto。
+# 避免 /api/v1/knowledge 等 redirect 在 HTTPS 反向代理後產生 http:// Location。
 app.add_middleware(ProxyHeadersMiddleware, trusted_hosts="*")
 
 # CORS - Strict Whitelist (Iron Law #2)
@@ -1006,6 +1054,15 @@ app.include_router(platform_v1.router, prefix="/api/v1/platform", tags=["AwoooP
 @app.get("/metrics", include_in_schema=False)
 async def prometheus_metrics() -> Response:
     """Prometheus metrics endpoint for alerting"""
+    # 2026-05-19 Codex — T85 Alert Chain DB evidence refresh.
+    # record_alert_chain_success() 是 process-local gauge；部署後第一個 scrape
+    # 可能尚未收到新 webhook，導致 smoke test 誤判 metric 不存在。
+    # 先用 AwoooP inbound / alert_operation_log 的 durable evidence 回填 last_success。
+    try:
+        await get_alert_chain_metrics_service().refresh_last_success_gauge()
+    except Exception as exc:
+        logger.warning("prometheus_metrics_alert_chain_evidence_error", error=str(exc))
+
     content = generate_latest().decode("utf-8")
     # 2026-05-07 ogt + Claude Sonnet 4.6 — INC-20260507-99ADF2 修復
     # 飛輪指標（awoooi_flywheel_*）原本只在 /api/v1/stats/flywheel/metrics 暴露，
@@ -1016,6 +1073,13 @@ async def prometheus_metrics() -> Response:
         content += flywheel_metrics.to_prometheus_lines()
     except Exception:
         logger.warning("prometheus_metrics_flywheel_error")
+    # 2026-05-14 Codex — T18 ADR-100 SLO emitter
+    # GovernanceAgent 讀 Prometheus recording rules；若 /metrics 不吐底層 DB totals，
+    # sli:* rules 會全空並每小時重複發 governance_slo_data_gap。
+    try:
+        content += await get_adr100_slo_metrics_service().to_prometheus_lines()
+    except Exception as exc:
+        logger.warning("prometheus_metrics_adr100_slo_error", error=str(exc))
     return Response(content=content, media_type=CONTENT_TYPE_LATEST)
 
 

apps/api/src/models/approval.py

@@ -167,6 +167,8 @@ class ApprovalRequest(ApprovalRequestBase):
     fingerprint: str | None = Field(default=None, description="告警指紋 Hash")
     hit_count: int = Field(default=1, description="聚合觸發次數")
     last_seen_at: datetime = Field(default_factory=lambda: datetime.now(timezone.utc), description="最後觸發時間")
+    telegram_message_id: int | None = Field(default=None, description="Telegram approval card message ID")
+    telegram_chat_id: int | None = Field(default=None, description="Telegram chat ID for the approval card")
     # 2026-04-14 Claude Sonnet 4.6: incident_id 已移至 Base（避免 ApprovalRequestCreate 缺欄位）
 
     @property
@@ -216,6 +218,10 @@ class ApprovalRequestResponse(BaseModel):
     hit_count: int = 1
     last_seen_at: datetime | None = None
     # Phase 6.5: Incident 關聯 (用於簽核後更新 Incident 狀態)
+    incident_id: str | None = None
+    matched_playbook_id: str | None = None
+    telegram_message_id: int | None = None
+    telegram_chat_id: int | None = None
     metadata: dict | None = None
 
     @classmethod
@@ -241,6 +247,10 @@ class ApprovalRequestResponse(BaseModel):
             hit_count=approval.hit_count,
             last_seen_at=approval.last_seen_at,
             # Phase 6.5
+            incident_id=approval.incident_id,
+            matched_playbook_id=approval.matched_playbook_id,
+            telegram_message_id=approval.telegram_message_id,
+            telegram_chat_id=approval.telegram_chat_id,
             metadata=approval.metadata,
         )
 

apps/api/src/models/governance.py

@@ -87,13 +87,27 @@ class DispatchItem(BaseModel):
     governance_event_id: str
     event_type: str
     dispatch_status: str
+    executor_type: str | None = None
     proposed_action: str = Field(description="≤120 字動作摘要")
     playbook_id: str | None = None
     playbook_trust: float | None = Field(default=None, ge=0.0, le=1.0)
     created_at: datetime
     dispatched_at: datetime | None = None
+    started_at: datetime | None = None
     completed_at: datetime | None = None
     operator_note: str | None = None
+    decision_path: str | None = None
+    workflow_stage: str | None = None
+    workflow_steps: list[str] = Field(default_factory=list)
+    next_action: str | None = None
+    lead_agent: str | None = None
+    support_agents: list[str] = Field(default_factory=list)
+    human_owner: str | None = None
+    kb_draft_entry_id: str | None = None
+    worker_status: str | None = None
+    dry_run_plan_fingerprint: str | None = None
+    archived_count: int | None = None
+    stale_ratio_snapshot: dict | None = None
 
 
 class GovernanceQueueResponse(BaseModel):
@@ -107,6 +121,475 @@ class GovernanceQueueResponse(BaseModel):
     )
 
 
+# =============================================================================
+# Endpoint 2B: KM review draft dedupe
+# =============================================================================
+
+class KnowledgeReviewDraftDedupeGroup(BaseModel):
+    governance_event_id: str
+    canonical_entry_id: str
+    canonical_title: str
+    canonical_updated_at: datetime | None = None
+    preferred_source: Literal["dispatch_context", "latest_review_draft"]
+    duplicate_entry_ids: list[str] = Field(default_factory=list)
+    duplicate_count: int
+    total_entries: int
+    suggested_action: str
+    owner_action: str
+    writes_on_read: bool = False
+    can_archive_without_owner_approval: bool = False
+    archive_history: list[DispatchItem] = Field(default_factory=list)
+
+
+class KnowledgeReviewDraftDedupeResponse(BaseModel):
+    schema_version: str = "km_review_draft_dedupe_v1"
+    total_review_drafts: int
+    event_group_total: int
+    duplicate_draft_total: int
+    groups: list[KnowledgeReviewDraftDedupeGroup]
+    generated_at: datetime
+
+
+class KnowledgeReviewDraftArchiveRequest(BaseModel):
+    canonical_entry_id: str = Field(min_length=1, max_length=120)
+    duplicate_entry_ids: list[str] = Field(min_length=1, max_length=100)
+    owner: str = Field(default="operator_console", min_length=1, max_length=100)
+    owner_approved: bool = False
+    dry_run: bool = False
+    dry_run_plan_fingerprint: str | None = Field(
+        default=None,
+        max_length=80,
+        description="Dry-run response fingerprint that must be echoed before a write.",
+    )
+
+
+class KnowledgeReviewDraftStaleRatioSnapshot(BaseModel):
+    stale_count: int
+    total_count: int
+    stale_ratio: float
+    threshold: float
+    stale_days: int
+
+
+class KnowledgeReviewDraftArchiveResponse(BaseModel):
+    schema_version: str = "km_review_draft_archive_v1"
+    governance_event_id: str
+    canonical_entry_id: str
+    requested_duplicate_entry_ids: list[str]
+    archived_entry_ids: list[str] = Field(default_factory=list)
+    skipped_entry_ids: list[str] = Field(default_factory=list)
+    would_archive_entry_ids: list[str] = Field(default_factory=list)
+    status: Literal["dry_run", "archived", "noop_already_archived"]
+    owner: str
+    owner_approved: bool
+    dry_run: bool
+    writes_km: bool
+    writes_governance_audit: bool
+    audit_dispatch_id: str | None = None
+    stale_ratio_snapshot: KnowledgeReviewDraftStaleRatioSnapshot | None = None
+    stale_ratio_recheck_status: Literal[
+        "dry_run",
+        "completed",
+        "already_active",
+        "not_requested",
+    ] = "not_requested"
+    stale_ratio_recheck_dispatch_id: str | None = None
+    dry_run_plan_fingerprint: str | None = None
+    next_action: str = "stale_ratio_recheck"
+    generated_at: datetime
+
+
+# =============================================================================
+# Endpoint 2C: KM stale candidates
+# =============================================================================
+
+class KnowledgeStaleCandidate(BaseModel):
+    entry_id: str
+    project_id: str
+    title: str
+    entry_type: str
+    category: str | None = None
+    status: str
+    source: str | None = None
+    updated_at: datetime | None = None
+    stale_days: int
+    view_count: int
+    priority_score: int
+    priority_tier: Literal["P0", "P1", "P2"]
+    recommended_action: Literal[
+        "refresh_with_evidence",
+        "owner_review",
+        "archive_or_supersede",
+    ]
+    reasons: list[str] = Field(default_factory=list)
+    correlation_sources: list[str] = Field(default_factory=list)
+    related_incident_id: str | None = None
+    related_playbook_id: str | None = None
+    related_approval_id: str | None = None
+    tags: list[str] = Field(default_factory=list)
+    owner_review_dispatch_id: str | None = None
+    owner_review_status: str | None = None
+    owner_review_stage: str | None = None
+    owner_review_next_action: str | None = None
+
+
+class KnowledgeStaleCandidatesResponse(BaseModel):
+    schema_version: str = "km_stale_candidates_v1"
+    project_id: str
+    total_stale: int
+    returned: int
+    threshold_days: int
+    writes_on_read: bool = False
+    manual_review_required: bool = True
+    items: list[KnowledgeStaleCandidate]
+    generated_at: datetime
+
+
+class KnowledgeStaleOwnerReviewRequest(BaseModel):
+    owner: str = Field(default="operator_console", min_length=1, max_length=100)
+    owner_note: str | None = Field(default=None, max_length=240)
+    dry_run: bool = False
+
+
+class KnowledgeStaleOwnerReviewResponse(BaseModel):
+    schema_version: str = "km_stale_owner_review_v1"
+    entry_id: str
+    project_id: str
+    status: Literal["dry_run", "queued", "already_queued"]
+    governance_event_id: str | None = None
+    dispatch_id: str | None = None
+    workflow_stage: str
+    recommended_action: Literal[
+        "refresh_with_evidence",
+        "owner_review",
+        "archive_or_supersede",
+    ]
+    owner: str
+    owner_note: str | None = None
+    writes_km: bool = False
+    writes_governance_audit: bool
+    next_action: str = "owner_review_stale_km_candidate"
+    generated_at: datetime
+
+
+class KnowledgeStaleOwnerReviewBatchQueueRequest(BaseModel):
+    project_id: str = Field(default="awoooi", min_length=1, max_length=64)
+    priority_tiers: list[Literal["P0", "P1", "P2"]] = Field(
+        default_factory=lambda: ["P0", "P1"],
+        min_length=1,
+        max_length=3,
+    )
+    limit: int = Field(default=10, ge=1, le=50)
+    owner: str = Field(default="operator_console", min_length=1, max_length=100)
+    owner_note: str | None = Field(default=None, max_length=240)
+    dry_run: bool = False
+    dry_run_plan_fingerprint: str | None = Field(
+        default=None,
+        max_length=80,
+        description="Dry-run response fingerprint that must be echoed before queueing a batch.",
+    )
+
+
+class KnowledgeStaleOwnerReviewBatchItem(BaseModel):
+    entry_id: str
+    title: str
+    priority_tier: Literal["P0", "P1", "P2"]
+    recommended_action: Literal[
+        "refresh_with_evidence",
+        "owner_review",
+        "archive_or_supersede",
+    ]
+    status: Literal["would_queue", "queued", "already_queued", "skipped"]
+    reason: str | None = None
+    governance_event_id: str | None = None
+    dispatch_id: str | None = None
+    workflow_stage: str
+
+
+class KnowledgeStaleOwnerReviewBatchQueueResponse(BaseModel):
+    schema_version: str = "km_stale_owner_review_batch_v1"
+    project_id: str
+    status: Literal["dry_run", "queued", "noop_already_queued"]
+    owner: str
+    owner_note: str | None = None
+    dry_run: bool
+    priority_tiers: list[str]
+    requested_limit: int
+    candidate_count: int
+    queued_count: int
+    already_queued_count: int
+    skipped_count: int
+    batch_governance_event_id: str | None = None
+    batch_dispatch_id: str | None = None
+    workflow_stage: str
+    writes_km: bool = False
+    writes_governance_audit: bool
+    stale_ratio_snapshot: KnowledgeReviewDraftStaleRatioSnapshot | None = None
+    dry_run_plan_fingerprint: str | None = None
+    items: list[KnowledgeStaleOwnerReviewBatchItem] = Field(default_factory=list)
+    next_action: str = "owner_review_stale_km_batch"
+    generated_at: datetime
+
+
+class KnowledgeStaleOwnerReviewInboxItem(BaseModel):
+    dispatch_id: str
+    governance_event_id: str
+    entry_id: str
+    project_id: str
+    title: str
+    dispatch_status: str
+    workflow_stage: str
+    next_action: str | None = None
+    owner: str | None = None
+    owner_note: str | None = None
+    batch_governance_event_id: str | None = None
+    batch_dispatch_id: str | None = None
+    priority_tier: Literal["P0", "P1", "P2"]
+    priority_score: int
+    recommended_action: Literal[
+        "refresh_with_evidence",
+        "owner_review",
+        "archive_or_supersede",
+    ]
+    stale_days: int
+    view_count: int
+    correlation_sources: list[str] = Field(default_factory=list)
+    reasons: list[str] = Field(default_factory=list)
+    related_incident_id: str | None = None
+    related_playbook_id: str | None = None
+    related_approval_id: str | None = None
+    dry_run_plan_fingerprint: str | None = None
+    queued_at: datetime | None = None
+    started_at: datetime | None = None
+    completed_at: datetime | None = None
+
+
+class KnowledgeStaleOwnerReviewInboxResponse(BaseModel):
+    schema_version: str = "km_stale_owner_review_inbox_v1"
+    project_id: str
+    dispatch_status: str
+    total: int
+    returned: int
+    writes_on_read: bool = False
+    manual_review_required: bool = True
+    items: list[KnowledgeStaleOwnerReviewInboxItem] = Field(default_factory=list)
+    generated_at: datetime
+
+
+class KnowledgeStaleOwnerReviewBurnDownItem(BaseModel):
+    completion_dispatch_id: str
+    governance_event_id: str
+    source_dispatch_id: str | None = None
+    recheck_dispatch_id: str | None = None
+    entry_id: str | None = None
+    project_id: str
+    dispatch_status: str
+    workflow_stage: str
+    review_outcome: Literal[
+        "refresh_with_evidence",
+        "archive",
+        "supersede",
+    ] | None = None
+    owner: str | None = None
+    completed_at: datetime | None = None
+    stale_ratio_snapshot: KnowledgeReviewDraftStaleRatioSnapshot | None = None
+    stale_count_delta: int | None = None
+    stale_ratio_delta: float | None = None
+    above_threshold: bool | None = None
+
+
+class KnowledgeStaleOwnerReviewBurnDownResponse(BaseModel):
+    schema_version: str = "km_stale_owner_review_burndown_v1"
+    project_id: str
+    burn_down_status: Literal["above_threshold", "at_or_below_threshold", "no_data"]
+    current_snapshot: KnowledgeReviewDraftStaleRatioSnapshot | None = None
+    entries_to_threshold: int
+    pending_owner_reviews: int
+    completed_owner_reviews: int
+    completion_audit_total: int
+    stale_ratio_recheck_total: int
+    latest_stale_count_delta: int | None = None
+    latest_stale_ratio_delta: float | None = None
+    writes_on_read: bool = False
+    manual_review_required: bool = True
+    returned: int
+    items: list[KnowledgeStaleOwnerReviewBurnDownItem] = Field(default_factory=list)
+    generated_at: datetime
+
+
+class KnowledgeStaleOwnerReviewCompletionQueueItem(BaseModel):
+    dispatch_id: str
+    governance_event_id: str
+    entry_id: str
+    project_id: str
+    title: str
+    dispatch_status: str
+    workflow_stage: str
+    readiness: Literal["ready", "blocked", "completed", "failed"]
+    recommended_completion_outcome: Literal[
+        "refresh_with_evidence",
+        "archive",
+        "supersede",
+    ]
+    next_action: str
+    blockers: list[str] = Field(default_factory=list)
+    required_owner_fields: list[str] = Field(default_factory=list)
+    can_preview: bool
+    can_confirm_after_preview: bool
+    writes_km_on_confirm: bool
+    owner: str | None = None
+    owner_note: str | None = None
+    batch_governance_event_id: str | None = None
+    batch_dispatch_id: str | None = None
+    priority_tier: Literal["P0", "P1", "P2"]
+    priority_score: int
+    recommended_action: Literal[
+        "refresh_with_evidence",
+        "owner_review",
+        "archive_or_supersede",
+    ]
+    stale_days: int
+    view_count: int
+    correlation_sources: list[str] = Field(default_factory=list)
+    reasons: list[str] = Field(default_factory=list)
+    related_incident_id: str | None = None
+    related_playbook_id: str | None = None
+    related_approval_id: str | None = None
+    dry_run_plan_fingerprint: str | None = None
+    queued_at: datetime | None = None
+    started_at: datetime | None = None
+    completed_at: datetime | None = None
+
+
+class KnowledgeStaleOwnerReviewCompletionQueueResponse(BaseModel):
+    schema_version: str = "km_stale_owner_review_completion_queue_v1"
+    project_id: str
+    status_bucket: Literal["all", "ready", "blocked", "completed", "failed", "pending"]
+    priority_tiers: list[str] = Field(default_factory=list)
+    recommended_completion_outcome: Literal[
+        "all",
+        "refresh_with_evidence",
+        "archive",
+        "supersede",
+    ] = "all"
+    batch_governance_event_id: str | None = None
+    can_preview: bool | None = None
+    total: int
+    returned: int
+    pending_count: int
+    ready_count: int
+    blocked_count: int
+    completed_count: int
+    failed_count: int
+    writes_on_read: bool = False
+    manual_review_required: bool = True
+    batch_writes_allowed: bool = False
+    items: list[KnowledgeStaleOwnerReviewCompletionQueueItem] = Field(default_factory=list)
+    generated_at: datetime
+
+
+class KnowledgeStaleOwnerReviewCompletionBatchPreviewRequest(BaseModel):
+    project_id: str = Field(default="awoooi", min_length=1, max_length=64)
+    status_bucket: Literal["all", "ready", "blocked", "completed", "failed", "pending"] = "ready"
+    priority_tiers: list[Literal["P0", "P1", "P2"]] = Field(
+        default_factory=lambda: ["P0", "P1", "P2"],
+        min_length=1,
+        max_length=3,
+    )
+    recommended_completion_outcome: Literal[
+        "all",
+        "refresh_with_evidence",
+        "archive",
+        "supersede",
+    ] = "all"
+    batch_governance_event_id: str | None = Field(default=None, max_length=120)
+    limit: int = Field(default=10, ge=1, le=30)
+    owner: str = Field(default="operator_console", min_length=1, max_length=100)
+    owner_note: str | None = Field(default=None, max_length=240)
+
+
+class KnowledgeStaleOwnerReviewCompletionBatchPreviewResponse(BaseModel):
+    schema_version: str = "km_stale_owner_review_completion_batch_preview_v1"
+    project_id: str
+    status: Literal["dry_run"] = "dry_run"
+    owner: str
+    owner_note: str | None = None
+    status_bucket: Literal["all", "ready", "blocked", "completed", "failed", "pending"]
+    priority_tiers: list[str]
+    recommended_completion_outcome: Literal[
+        "all",
+        "refresh_with_evidence",
+        "archive",
+        "supersede",
+    ]
+    batch_governance_event_id: str | None = None
+    requested_limit: int
+    candidate_count: int
+    previewable_count: int
+    blocked_count: int
+    completed_count: int
+    failed_count: int
+    writes_km: bool = False
+    writes_governance_audit: bool = False
+    batch_writes_allowed: bool = False
+    manual_review_required: bool = True
+    dry_run_plan_fingerprint: str
+    next_action: str = "preview_each_ready_item_then_confirm_single_item"
+    items: list[KnowledgeStaleOwnerReviewCompletionQueueItem] = Field(default_factory=list)
+    generated_at: datetime
+
+
+class KnowledgeStaleOwnerReviewCompleteRequest(BaseModel):
+    dispatch_id: str | None = Field(
+        default=None,
+        max_length=120,
+        description="Owner-review dispatch id. Optional when the backend can resolve the active item by entry id.",
+    )
+    owner: str = Field(default="operator_console", min_length=1, max_length=100)
+    owner_approved: bool = False
+    dry_run: bool = False
+    review_outcome: Literal[
+        "refresh_with_evidence",
+        "archive",
+        "supersede",
+    ]
+    owner_note: str | None = Field(default=None, max_length=500)
+    updated_title: str | None = Field(default=None, min_length=1, max_length=255)
+    updated_content: str | None = Field(default=None, min_length=1)
+    superseded_by_entry_id: str | None = Field(default=None, max_length=120)
+    dry_run_plan_fingerprint: str | None = Field(
+        default=None,
+        max_length=80,
+        description="Dry-run response fingerprint that must be echoed before a write.",
+    )
+
+
+class KnowledgeStaleOwnerReviewCompleteResponse(BaseModel):
+    schema_version: str = "km_stale_owner_review_complete_v1"
+    entry_id: str
+    project_id: str
+    status: Literal["dry_run", "completed", "already_completed"]
+    review_outcome: Literal[
+        "refresh_with_evidence",
+        "archive",
+        "supersede",
+    ]
+    governance_event_id: str
+    dispatch_id: str
+    audit_dispatch_id: str | None = None
+    stale_ratio_recheck_dispatch_id: str | None = None
+    workflow_stage: str
+    owner: str
+    owner_approved: bool
+    dry_run: bool
+    writes_km: bool
+    writes_governance_audit: bool
+    stale_ratio_snapshot: KnowledgeReviewDraftStaleRatioSnapshot | None = None
+    dry_run_plan_fingerprint: str | None = None
+    next_action: str = "stale_ratio_recheck"
+    generated_at: datetime
+
+
 # =============================================================================
 # Endpoint 3: summary
 # =============================================================================

apps/api/src/plugins/mcp/gateway.py

@@ -388,10 +388,7 @@ class McpGateway:
         parameters: dict[str, Any],
     ) -> MCPToolResult:
         """呼叫底層 MCP provider 執行工具"""
-        registry = get_provider_registry()
-        provider = registry.get(ctx.tool_name) or registry.get(
-            tool_row.tool_name if tool_row else ctx.tool_name
-        )
+        provider = await self._resolve_provider(ctx, tool_row)
 
         # 找不到 provider → 回傳 shadow no-op
         if provider is None:
@@ -407,14 +404,57 @@ class McpGateway:
             )
 
         audit_params = dict(parameters)
+        existing_audit = (
+            parameters.get("_mcp_audit")
+            if isinstance(parameters, dict) and isinstance(parameters.get("_mcp_audit"), dict)
+            else {}
+        )
         audit_params["_mcp_audit"] = {
             "project_id": ctx.project_id,
             "agent_id": ctx.agent_id,
             "run_id": str(ctx.run_id) if ctx.run_id else None,
             "trace_id": ctx.trace_id,
+            "incident_id": existing_audit.get("incident_id") or ctx.trace_id,
+            "session_id": existing_audit.get("session_id"),
+            "flywheel_node": existing_audit.get("flywheel_node"),
+            "agent_role": existing_audit.get("agent_role") or ctx.agent_id,
+            "gateway_path": "awooop_mcp_gateway",
         }
         return await provider.execute(ctx.tool_name, audit_params)
 
+    async def _resolve_provider(
+        self,
+        ctx: GatewayContext,
+        tool_row: AwoooPMcpToolRegistry | None,
+    ):
+        """Find the provider that owns ctx.tool_name.
+
+        ProviderRegistry is keyed by provider name (`kubernetes`, `ssh_host`, ...),
+        while GatewayContext intentionally uses the governed tool name
+        (`kubectl_get`, `ssh_diagnose`, ...).  Scan provider tool manifests as the
+        compatibility bridge until registry exposes a first-class tool index.
+        """
+        registry = get_provider_registry()
+        direct = registry.get(ctx.tool_name)
+        if direct is not None:
+            return direct
+
+        lookup_name = tool_row.tool_name if tool_row else ctx.tool_name
+        for provider in registry.all():
+            try:
+                tools = await provider.list_tools()
+            except Exception as exc:
+                logger.debug(
+                    "mcp_gateway_provider_manifest_skipped",
+                    provider=getattr(provider, "name", None),
+                    tool_name=lookup_name,
+                    error=str(exc),
+                )
+                continue
+            if any(tool.name == lookup_name for tool in tools):
+                return provider
+        return None
+
     # ── Audit log ─────────────────────────────────────────────────────────────
 
     async def _write_audit(
@@ -442,6 +482,15 @@ class McpGateway:
                     json.dumps(result.output, sort_keys=True, default=str).encode()
                 ).hexdigest()
 
+            gate_payload = {
+                **gate_result.as_dict(),
+                "schema_version": "awooop_mcp_gateway_audit_v1",
+                "gateway_path": "awooop_mcp_gateway",
+                "policy_enforced": True,
+                "is_shadow": ctx.is_shadow,
+                "required_scope": ctx.required_scope,
+            }
+
             audit = AwoooPMcpGatewayAudit(
                 project_id=ctx.project_id,
                 run_id=ctx.run_id,
@@ -451,7 +500,7 @@ class McpGateway:
                 tool_name=ctx.tool_name,
                 input_hash=input_hash,
                 output_hash=output_hash,
-                gate_result=gate_result.as_dict(),
+                gate_result=gate_payload,
                 result_status=result_status,
                 block_gate=block_gate,
                 block_reason=block_reason,

apps/api/src/plugins/mcp/providers/ssh_provider.py

@@ -65,6 +65,13 @@ DEFAULT_HOST_USERS = {
     # AI/Web host is operated by the ollama account in the current topology.
     "192.168.0.188": "ollama",
 }
+SHORT_HOST_MAP = {
+    "110": "192.168.0.110",
+    "120": "192.168.0.120",
+    "121": "192.168.0.121",
+    "188": "192.168.0.188",
+    "wooo": "192.168.0.110",
+}
 DIAG_TIMEOUT = 10   # 診斷類超時（秒）
 OP_TIMEOUT   = 60   # 操作類超時（秒）
 
@@ -127,7 +134,9 @@ def _normalize_ssh_host(value: str) -> str:
     if host.count(":") == 1:
         maybe_host, maybe_port = host.rsplit(":", 1)
         if maybe_port.isdigit():
-            return maybe_host
+            host = maybe_host
+    if host in SHORT_HOST_MAP:
+        return SHORT_HOST_MAP[host]
     return host
 
 
@@ -240,6 +249,10 @@ class SSHProvider(MCPToolProvider):
                 ),
                 input_schema={"type": "object", "properties": {
                     "host": {"type": "string", "description": "Target host IP"},
+                    "container_name": {
+                        "type": "string",
+                        "description": "Optional Docker container name for container-focused diagnostics",
+                    },
                 }, "required": ["host"]},
                 server_name=self.name,
             ),
@@ -542,12 +555,23 @@ class SSHProvider(MCPToolProvider):
         # 所有接受用戶字串的工具，必須先通過 _validate_param() 白名單驗證
         if tool_name == "ssh_diagnose":
             # 2026-04-27 Claude Sonnet 4.6: 主機告警自動診斷 — 只讀，不修改任何狀態
-            return (
+            command = (
                 "echo '=== CPU TOP ===' && ps aux --sort=-%cpu | head -15 && "
                 "echo '=== MEMORY ===' && free -h && "
                 "echo '=== DISK ===' && df -h && "
                 "echo '=== LOAD ===' && uptime"
             )
+            container_name = params.get("container_name")
+            if container_name:
+                name = _validate_param("container_name", str(container_name))
+                command = (
+                    f"{command} && "
+                    f"echo '=== DOCKER STATS {name} ===' && "
+                    f"docker stats --no-stream {name} 2>&1 && "
+                    f"echo '=== DOCKER INSPECT {name} ===' && "
+                    f"docker inspect {name} 2>&1 | head -80"
+                )
+            return command
 
         if tool_name == "ssh_get_top_processes":
             return "ps aux --sort=-%cpu | head -15"
@@ -564,7 +588,10 @@ class SSHProvider(MCPToolProvider):
             return f"docker logs {name} --tail {tail} 2>&1"
 
         if tool_name == "ssh_get_container_status":
-            name = _validate_param("filter_name", params["filter_name"])
+            raw_name = params.get("filter_name") or params.get("container_name") or params.get("name")
+            if not raw_name:
+                raise ValueError("Missing filter_name for ssh_get_container_status")
+            name = _validate_param("filter_name", str(raw_name))
             return f"docker ps -a --filter name={name}"
 
         if tool_name == "ssh_get_service_status":

apps/api/src/repositories/approval_repository.py

@@ -106,6 +106,8 @@ def _record_to_request(record: ApprovalRecord) -> ApprovalRequest:
         # B4 fix 2026-04-24 ogt + Claude Sonnet 4.6: 補回 DB 欄位（人工審核路徑讀回必要）
         incident_id=getattr(record, "incident_id", None),
         matched_playbook_id=getattr(record, "matched_playbook_id", None),
+        telegram_message_id=getattr(record, "telegram_message_id", None),
+        telegram_chat_id=getattr(record, "telegram_chat_id", None),
     )
 
 

apps/api/src/repositories/drift_repository.py

@@ -18,7 +18,14 @@ import structlog
 from sqlalchemy import text
 
 from src.db.base import get_db_context
-from src.models.drift import DriftInterpretation, DriftIntent, DriftItem, DriftLevel, DriftReport, DriftStatus
+from src.models.drift import (
+    DriftIntent,
+    DriftInterpretation,
+    DriftItem,
+    DriftLevel,
+    DriftReport,
+    DriftStatus,
+)
 
 logger = structlog.get_logger(__name__)
 
@@ -167,6 +174,40 @@ class DriftReportRepository:
                 {"report_id": report_id, "narrative": narrative},
             )
 
+    async def get_repeat_state(
+        self,
+        report: DriftReport,
+        *,
+        include_values: bool = True,
+    ) -> dict:
+        """Return stable fingerprint repeat state for a drift report."""
+        from src.services.drift_repeat_state import build_drift_repeat_state
+
+        async with get_db_context() as db:
+            result = await db.execute(
+                text("""
+                    SELECT
+                        report_id,
+                        namespace,
+                        status,
+                        scanned_at,
+                        created_at,
+                        items
+                    FROM drift_reports
+                    WHERE namespace = :namespace
+                      AND created_at > now() - interval '24 hours'
+                    ORDER BY scanned_at DESC
+                    LIMIT 200
+                """),
+                {"namespace": report.namespace},
+            )
+            rows = [dict(row) for row in result.mappings().all()]
+        return build_drift_repeat_state(
+            report,
+            rows,
+            include_values=include_values,
+        )
+
 
 _drift_repo: DriftReportRepository | None = None
 

apps/api/src/repositories/governance_remediation_dispatch_repo.py

@@ -356,6 +356,75 @@ async def list_pending(
         return list(result.scalars().all())
 
 
+async def list_pending_by_executor(
+    executor_type: str,
+    *,
+    limit: int = 50,
+) -> list[GovernanceRemediationDispatch]:
+    """列出指定 executor 的 pending dispatch（按 dispatched_at ASC）。
+
+    用於 Hermes / 其他 worker 消費自己的 work item。由 repository 層集中查詢，
+    避免 job 直接散落表名與狀態條件。
+
+    Args:
+        executor_type: dispatch.executor_type，例如 hermes_kb_growth_healthcheck
+        limit: 本輪最多取幾筆，避免 backlog 一次拖垮 worker
+
+    Returns:
+        最舊優先的 pending dispatch 列表。
+    """
+    async with get_db_context() as db:
+        result = await db.execute(
+            select(GovernanceRemediationDispatch)
+            .where(GovernanceRemediationDispatch.dispatch_status == "pending")
+            .where(GovernanceRemediationDispatch.executor_type == executor_type)
+            .order_by(GovernanceRemediationDispatch.dispatched_at.asc())
+            .limit(limit)
+        )
+        return list(result.scalars().all())
+
+
+async def update_decision_context(
+    dispatch_id: str,
+    decision_context: dict[str, Any],
+) -> GovernanceRemediationDispatch:
+    """更新 dispatch 的 decision_context，保留同一 row 的 audit trail。
+
+    這只更新 dispatch work item 的讀模型上下文，不修改 immutable
+    ai_governance_events，也不代表治理事件已被解決。
+
+    Args:
+        dispatch_id: governance_remediation_dispatch.id
+        decision_context: 新的 JSONB context
+
+    Returns:
+        更新後的 GovernanceRemediationDispatch ORM 物件
+
+    Raises:
+        DispatchNotFound: 找不到 dispatch_id
+    """
+    async with get_db_context() as db:
+        result = await db.execute(
+            select(GovernanceRemediationDispatch)
+            .where(GovernanceRemediationDispatch.id == dispatch_id)
+        )
+        row = result.scalar_one_or_none()
+        if row is None:
+            raise DispatchNotFound(f"dispatch_id={dispatch_id!r} 不存在")
+
+        row.decision_context = decision_context
+        await db.flush()
+        await db.refresh(row)
+
+        logger.info(
+            "dispatch_decision_context_updated",
+            dispatch_id=dispatch_id,
+            event_id=row.governance_event_id,
+            executor_type=row.executor_type,
+        )
+        return row
+
+
 async def list_by_event(
     event_id: str,
 ) -> list[GovernanceRemediationDispatch]:

apps/api/src/repositories/incident_repository.py

@@ -19,7 +19,12 @@ from sqlalchemy import select
 
 from src.db.base import get_db_context
 from src.db.models import IncidentRecord
-from src.models.incident import Incident, IncidentFrequencyStats, IncidentStatus, Severity
+from src.models.incident import (
+    Incident,
+    IncidentFrequencyStats,
+    IncidentStatus,
+    Severity,
+)
 from src.repositories.interfaces import IIncidentRepository
 
 logger = structlog.get_logger(__name__)
@@ -41,8 +46,8 @@ def _record_to_incident(record: IncidentRecord) -> Incident:
 
     return Incident(
         incident_id=record.incident_id,
-        status=IncidentStatus(record.status),
-        severity=Severity(record.severity),
+        status=IncidentStatus(_normalize_status(record.status)),
+        severity=Severity(_normalize_severity(record.severity)),
         signals=record.signals or [],
         affected_services=record.affected_services or [],
         proposal_ids=record.proposal_ids or [],
@@ -93,6 +98,36 @@ def _incident_to_record_data(incident: Incident) -> dict[str, Any]:
     }
 
 
+def _normalize_status(value: str | IncidentStatus) -> str:
+    if isinstance(value, IncidentStatus):
+        return value.value
+    raw = str(value)
+    if raw in IncidentStatus.__members__:
+        return IncidentStatus[raw].value
+    normalized = raw.strip().lower()
+    if normalized == "open":
+        return IncidentStatus.INVESTIGATING.value
+    return normalized
+
+
+def _normalize_severity(value: str | Severity) -> str:
+    if isinstance(value, Severity):
+        return value.value
+    raw = str(value)
+    if raw in Severity.__members__:
+        return Severity[raw].value
+    legacy_map = {
+        "critical": Severity.P0.value,
+        "high": Severity.P1.value,
+        "warning": Severity.P2.value,
+        "medium": Severity.P2.value,
+        "info": Severity.P3.value,
+        "low": Severity.P3.value,
+        "none": Severity.P3.value,
+    }
+    return legacy_map.get(raw.strip().lower(), raw)
+
+
 # =============================================================================
 # IncidentDBRepository
 # =============================================================================
@@ -136,8 +171,8 @@ class IncidentDBRepository(IIncidentRepository):
     async def get_active(self) -> list[Incident]:
         """取得所有活躍的 Incident"""
         active_statuses = [
-            IncidentStatus.INVESTIGATING.value,
-            IncidentStatus.MITIGATING.value,
+            IncidentStatus.INVESTIGATING,
+            IncidentStatus.MITIGATING,
         ]
         async with get_db_context() as db:
             result = await db.execute(

apps/api/src/repositories/knowledge_repository.py

@@ -190,7 +190,7 @@ class KnowledgeDBRepository:
             count_query = count_query.where(KnowledgeEntryRecord.status == status)
         if tags:
             for tag in tags:
-                tag_filter = KnowledgeEntryRecord.tags.op('@>')(f'["{tag}"]')
+                tag_filter = _json_string_array_has_tag(tag)
                 query = query.where(tag_filter)
                 count_query = count_query.where(tag_filter)
         if q:
@@ -347,3 +347,18 @@ class KnowledgeDBRepository:
             created_at=record.created_at,
             updated_at=record.updated_at,
         )
+
+
+def _json_string_array_has_tag(tag: str):
+    """建立 JSON/JSONB 皆相容的 tag filter。
+
+    production 的 knowledge_entries.tags 目前是 JSON 欄位，不支援 json @> text。
+    這裡改用帶引號的字串比對，避免把 tag 片段誤判成完整 tag。
+    """
+    escaped = (
+        tag
+        .replace("\\", "\\\\")
+        .replace("%", "\\%")
+        .replace("_", "\\_")
+    )
+    return KnowledgeEntryRecord.tags.cast(String).ilike(f'%"{escaped}"%', escape="\\")

apps/api/src/routes/agent.py

@@ -19,10 +19,11 @@ router = APIRouter()
 logger = logging.getLogger(__name__)
 
 # ==================== Ollama Config ====================
-# 2026-05-03 ogt: ADR-110 GCP-A Primary — 改從 settings 讀取，不再硬編碼 111
-def _get_ollama_base_url() -> str:
-    from src.core.config import get_settings
-    return get_settings().OLLAMA_URL
+# 2026-05-19 Codex: agent thinking stream follows GCP-A → GCP-B → 111.
+def _get_ollama_endpoints():
+    from src.services.ollama_endpoint_resolver import resolve_ollama_order
+
+    return resolve_ollama_order("interactive")
 OLLAMA_MODEL = "llama3.2:latest"  # 可根據實際部署調整
 OLLAMA_TIMEOUT = 120.0  # 串流超時
 
@@ -112,66 +113,82 @@ async def get_agent_thinking(
         # 1. 開始思考
         yield f"data: {json.dumps({'type': 'thinking', 'content': '正在連接 AI 模型...'}, ensure_ascii=False)}\n\n"
 
-        try:
-            async with httpx.AsyncClient(timeout=OLLAMA_TIMEOUT) as client:
-                # 2. 發送請求到 Ollama
-                yield f"data: {json.dumps({'type': 'thinking', 'content': f'模型: {model}'}, ensure_ascii=False)}\n\n"
+        last_error = ""
+        async with httpx.AsyncClient(timeout=OLLAMA_TIMEOUT) as client:
+            # 2. 發送請求到 Ollama
+            yield f"data: {json.dumps({'type': 'thinking', 'content': f'模型: {model}'}, ensure_ascii=False)}\n\n"
 
-                async with client.stream(
-                    "POST",
-                    f"{_get_ollama_base_url()}/api/generate",
-                    json={
-                        "model": model,
-                        "prompt": prompt,
-                        "stream": True,
-                    },
-                ) as response:
-                    if response.status_code != 200:
-                        yield f"data: {json.dumps({'type': 'error', 'content': f'Ollama 錯誤: HTTP {response.status_code}'}, ensure_ascii=False)}\n\n"
-                        yield "data: [DONE]\n\n"
-                        return
-
-                    yield f"data: {json.dumps({'type': 'thinking', 'content': '開始接收 AI 回應...'}, ensure_ascii=False)}\n\n"
-
-                    # 3. 串流讀取 Ollama 回應
-                    buffer = ""
-                    async for line in response.aiter_lines():
-                        if not line:
+            for endpoint in _get_ollama_endpoints():
+                if not endpoint.url:
+                    continue
+                try:
+                    async with client.stream(
+                        "POST",
+                        f"{endpoint.url}/api/generate",
+                        json={
+                            "model": model,
+                            "prompt": prompt,
+                            "stream": True,
+                        },
+                    ) as response:
+                        if response.status_code != 200:
+                            last_error = f"HTTP {response.status_code}"
+                            logger.warning(
+                                "agent_thinking_ollama_http_error",
+                                provider=endpoint.provider_name,
+                                status=response.status_code,
+                            )
                             continue
 
-                        try:
-                            chunk = json.loads(line)
-                            token = chunk.get("response", "")
-                            done = chunk.get("done", False)
+                        yield f"data: {json.dumps({'type': 'thinking', 'content': '開始接收 AI 回應...'}, ensure_ascii=False)}\n\n"
 
-                            if token:
-                                # 累積 token，每 10 字符或遇到標點符號時發送
-                                buffer += token
-                                if len(buffer) >= 10 or any(p in buffer for p in "。！？，、\n"):
-                                    yield f"data: {json.dumps({'type': 'thinking', 'content': buffer}, ensure_ascii=False)}\n\n"
-                                    buffer = ""
+                        # 3. 串流讀取 Ollama 回應
+                        buffer = ""
+                        async for line in response.aiter_lines():
+                            if not line:
+                                continue
 
-                            if done:
-                                # 發送剩餘 buffer
-                                if buffer:
-                                    yield f"data: {json.dumps({'type': 'thinking', 'content': buffer}, ensure_ascii=False)}\n\n"
-                                # 發送完成訊息
-                                yield f"data: {json.dumps({'type': 'result', 'content': '分析完成'}, ensure_ascii=False)}\n\n"
-                                break
+                            try:
+                                chunk = json.loads(line)
+                                token = chunk.get("response", "")
+                                done = chunk.get("done", False)
 
-                        except json.JSONDecodeError as e:
-                            logger.warning(f"JSON 解析失敗: {line[:100]}... - {e}")
-                            continue
+                                if token:
+                                    # 累積 token，每 10 字符或遇到標點符號時發送
+                                    buffer += token
+                                    if len(buffer) >= 10 or any(p in buffer for p in "。！？，、\n"):
+                                        yield f"data: {json.dumps({'type': 'thinking', 'content': buffer}, ensure_ascii=False)}\n\n"
+                                        buffer = ""
 
-        except httpx.ConnectError as e:
-            logger.error(f"無法連接 Ollama: {e}")
-            yield f"data: {json.dumps({'type': 'error', 'content': f'無法連接 Ollama ({_get_ollama_base_url()})'}, ensure_ascii=False)}\n\n"
-        except httpx.TimeoutException as e:
-            logger.error(f"Ollama 超時: {e}")
-            yield f"data: {json.dumps({'type': 'error', 'content': '請求超時'}, ensure_ascii=False)}\n\n"
-        except Exception as e:
-            logger.error(f"未知錯誤: {e}")
-            yield f"data: {json.dumps({'type': 'error', 'content': f'未知錯誤: {str(e)}'}, ensure_ascii=False)}\n\n"
+                                if done:
+                                    # 發送剩餘 buffer
+                                    if buffer:
+                                        yield f"data: {json.dumps({'type': 'thinking', 'content': buffer}, ensure_ascii=False)}\n\n"
+                                    # 發送完成訊息
+                                    yield f"data: {json.dumps({'type': 'result', 'content': '分析完成'}, ensure_ascii=False)}\n\n"
+                                    yield "data: [DONE]\n\n"
+                                    return
+
+                            except json.JSONDecodeError as e:
+                                logger.warning(f"JSON 解析失敗: {line[:100]}... - {e}")
+                                continue
+                except (httpx.ConnectError, httpx.TimeoutException) as e:
+                    last_error = type(e).__name__
+                    logger.error(
+                        "agent_thinking_ollama_endpoint_failed",
+                        provider=endpoint.provider_name,
+                        error=str(e),
+                    )
+                except Exception as e:
+                    last_error = str(e)
+                    logger.error(
+                        "agent_thinking_unknown_error",
+                        provider=endpoint.provider_name,
+                        error=str(e),
+                    )
+
+        error_content = f"Ollama 全端點不可用: {last_error or 'unknown'}"
+        yield f"data: {json.dumps({'type': 'error', 'content': error_content}, ensure_ascii=False)}\n\n"
 
         # 4. 結束標記
         yield "data: [DONE]\n\n"

apps/api/src/services/adr100_remediation_service.py

@@ -0,0 +1,606 @@
+"""
+ADR-100 Remediation Service
+===========================
+Safe operator entrypoints for verification remediation work items.
+
+T25: remediation queue items are now actionable without mutating incident state:
+- preview: show the selected guardrail path
+- dry-run: collect read-only current state and validate supported executor routing
+"""
+
+from __future__ import annotations
+
+import asyncio
+from typing import Any, Literal, Protocol
+
+import structlog
+
+from src.models.incident import Incident
+from src.repositories.incident_repository import IncidentDBRepository
+from src.services.adr100_slo_status_service import (
+    Adr100SloStatusService,
+    get_adr100_slo_status_service,
+)
+from src.services.auto_repair_service import AutoRepairService
+from src.services.post_execution_verifier import (
+    PostExecutionVerifier,
+    _assess_recovery,
+    _build_prometheus_query,
+    get_post_execution_verifier,
+)
+
+logger = structlog.get_logger(__name__)
+
+RemediationMode = Literal["auto", "reverify", "replay"]
+
+_READY_STATUSES = {"ready_for_replay", "ready_for_reverify"}
+
+
+class RemediationNotFoundError(LookupError):
+    """Requested ADR-100 remediation work item is not in the current read model."""
+
+
+class _IncidentRepository(Protocol):
+    async def get_by_id(self, incident_id: str) -> Incident | None:
+        ...
+
+
+class Adr100RemediationService:
+    """Read-only remediation preview and dry-run service."""
+
+    def __init__(
+        self,
+        *,
+        slo_service: Adr100SloStatusService | None = None,
+        incident_repository: _IncidentRepository | None = None,
+        auto_repair_service: AutoRepairService | None = None,
+        verifier: PostExecutionVerifier | None = None,
+        timeline_service: Any | None = None,
+        alert_operation_log_repository: Any | None = None,
+        record_history: bool = True,
+    ) -> None:
+        self._slo_service = slo_service or get_adr100_slo_status_service()
+        self._incident_repository = incident_repository or IncidentDBRepository()
+        self._auto_repair_service = auto_repair_service or AutoRepairService()
+        self._verifier = verifier or get_post_execution_verifier()
+        self._timeline_service = timeline_service
+        self._alert_operation_log_repository = alert_operation_log_repository
+        self._record_history_enabled = record_history
+
+    async def preview(self, work_item_id: str, mode: RemediationMode = "auto") -> dict[str, Any]:
+        """Return the safe execution plan for a remediation queue item."""
+
+        item = await self._find_work_item(work_item_id)
+        selected_mode = _select_mode(item, mode)
+        checks = _base_checks(item)
+        allowed = all(check["passed"] for check in checks)
+
+        return {
+            "schema_version": "adr100_remediation_preview_v1",
+            "work_item_id": item.get("work_item_id"),
+            "incident_id": item.get("incident_id"),
+            "auto_repair_id": item.get("auto_repair_id"),
+            "mode": selected_mode,
+            "allowed": allowed,
+            "safety_level": "read_only",
+            "writes_incident_state": False,
+            "writes_auto_repair_result": False,
+            "checks": checks,
+            "plan": _plan_for_item(item, selected_mode),
+            "source": "adr100.verification_coverage.remediation_queue",
+        }
+
+    async def dry_run(self, work_item_id: str, mode: RemediationMode = "auto") -> dict[str, Any]:
+        """Run a safe, read-only remediation dry-run for one queue item."""
+
+        item = await self._find_work_item(work_item_id)
+        selected_mode = _select_mode(item, mode)
+        checks = _base_checks(item)
+        incident = await self._load_incident(item)
+        checks.append({
+            "name": "incident_loaded",
+            "passed": incident is not None,
+            "detail": item.get("incident_id") or "missing incident_id",
+        })
+
+        if incident is None or not all(check["passed"] for check in checks):
+            payload = _dry_run_blocked_payload(item, selected_mode, checks)
+            payload["history"] = await self._record_dry_run_history(item, payload)
+            return payload
+
+        if selected_mode == "replay":
+            return await self._dry_run_replay(item, incident, checks)
+        return await self._dry_run_reverify(item, incident, checks)
+
+    async def history(
+        self,
+        *,
+        limit: int = 50,
+        incident_id: str | None = None,
+        work_item_id: str | None = None,
+    ) -> dict[str, Any]:
+        """Return durable dry-run history written by this remediation service."""
+
+        safe_limit = max(1, min(limit, 200))
+        fetch_limit = min(max(safe_limit * 4, 50), 200)
+        rows: list[Any] = []
+        repo = self._alert_operation_log_repository
+        if repo is None:
+            from src.repositories.alert_operation_log_repository import (
+                get_alert_operation_log_repository,
+            )
+
+            repo = get_alert_operation_log_repository()
+
+        for event_type in ("PRE_FLIGHT_PASSED", "PRE_FLIGHT_FAILED"):
+            try:
+                batch, _total = await repo.list_recent(
+                    limit=fetch_limit,
+                    event_type=event_type,
+                    incident_id=incident_id,
+                )
+                rows.extend(batch)
+            except Exception as exc:
+                logger.warning(
+                    "adr100_remediation_history_fetch_failed",
+                    event_type=event_type,
+                    incident_id=incident_id,
+                    error=str(exc),
+                )
+
+        rows.sort(key=_record_created_at, reverse=True)
+
+        items: list[dict[str, Any]] = []
+        for row in rows:
+            context = getattr(row, "context", None) or {}
+            if context.get("schema_version") != "adr100_remediation_dry_run_history_v1":
+                continue
+            if work_item_id and context.get("work_item_id") != work_item_id:
+                continue
+            items.append(_history_item(row, context))
+            if len(items) >= safe_limit:
+                break
+
+        return {
+            "schema_version": "adr100_remediation_history_v1",
+            "total": len(items),
+            "limit": safe_limit,
+            "filters": {
+                "incident_id": incident_id,
+                "work_item_id": work_item_id,
+            },
+            "items": items,
+            "by_work_item": _summarize_history_by_work_item(items),
+        }
+
+    async def _find_work_item(self, work_item_id: str) -> dict[str, Any]:
+        report = await self._slo_service.fetch_report()
+        coverage = report.get("verification_coverage") or {}
+        queue = coverage.get("remediation_queue") or {}
+
+        for item in queue.get("items") or []:
+            if item.get("work_item_id") == work_item_id:
+                return dict(item)
+
+        raise RemediationNotFoundError(work_item_id)
+
+    async def _load_incident(self, item: dict[str, Any]) -> Incident | None:
+        incident_id = str(item.get("incident_id") or "")
+        if not incident_id:
+            return None
+        return await self._incident_repository.get_by_id(incident_id)
+
+    async def _dry_run_reverify(
+        self,
+        item: dict[str, Any],
+        incident: Incident,
+        checks: list[dict[str, Any]],
+    ) -> dict[str, Any]:
+        post_state = await self._collect_current_state(incident)
+        action_taken = f"dry_run_reverify:{item.get('playbook_id') or 'unknown'}"
+        result = _assess_recovery(None, post_state, action_taken)
+
+        payload = _dry_run_result_payload(
+            item=item,
+            mode="reverify",
+            checks=checks,
+            post_state=post_state,
+            verification_result_preview=result,
+            extra={
+                "promql": _promql_for_incident(incident),
+                "mcp_route": {
+                    "agent_id": "post_execution_verifier",
+                    "required_scope": "read",
+                    "is_shadow": True,
+                    "flywheel_node": "verify",
+                },
+            },
+        )
+        payload["history"] = await self._record_dry_run_history(item, payload)
+        return payload
+
+    async def _dry_run_replay(
+        self,
+        item: dict[str, Any],
+        incident: Incident,
+        checks: list[dict[str, Any]],
+    ) -> dict[str, Any]:
+        diagnostic_command = _diagnostic_command_for_incident(incident)
+        route = self._auto_repair_service.preview_read_only_ssh_mcp_route(
+            incident,
+            diagnostic_command,
+        )
+        checks.append({
+            "name": "supported_executor_route",
+            "passed": route is not None,
+            "detail": "mcp:ssh_diagnose" if route else "missing host/container route",
+        })
+
+        post_state = await self._collect_current_state(incident)
+        action_taken = f"dry_run_replay:{item.get('playbook_id') or 'unknown'}"
+        result = _assess_recovery(None, post_state, action_taken)
+
+        payload = _dry_run_result_payload(
+            item=item,
+            mode="replay",
+            checks=checks,
+            post_state=post_state,
+            verification_result_preview=result,
+            extra={
+                "diagnostic_command_preview": diagnostic_command,
+                "mcp_route": route,
+                "promql": _promql_for_incident(incident),
+            },
+        )
+        payload["history"] = await self._record_dry_run_history(item, payload)
+        return payload
+
+    async def _collect_current_state(self, incident: Incident) -> dict[str, Any]:
+        try:
+            return await asyncio.wait_for(
+                self._verifier._collect_post_state(incident),
+                timeout=12.0,
+            )
+        except asyncio.TimeoutError:
+            logger.warning(
+                "adr100_remediation_dry_run_timeout",
+                incident_id=incident.incident_id,
+            )
+            return {}
+        except Exception as exc:
+            logger.warning(
+                "adr100_remediation_dry_run_collect_failed",
+                incident_id=incident.incident_id,
+                error=str(exc),
+            )
+            return {}
+
+    async def _record_dry_run_history(
+        self,
+        item: dict[str, Any],
+        payload: dict[str, Any],
+    ) -> dict[str, Any]:
+        if not self._record_history_enabled:
+            return {"recorded": False, "reason": "disabled"}
+
+        incident_id = str(item.get("incident_id") or "")
+        if not incident_id:
+            return {"recorded": False, "reason": "missing_incident_id"}
+
+        history: dict[str, Any] = {
+            "recorded": False,
+            "alert_operation_id": None,
+            "timeline_event_id": None,
+        }
+        context = _history_context(item, payload)
+        allowed = bool(payload.get("allowed"))
+
+        try:
+            repo = self._alert_operation_log_repository
+            if repo is None:
+                from src.repositories.alert_operation_log_repository import (
+                    get_alert_operation_log_repository,
+                )
+
+                repo = get_alert_operation_log_repository()
+            record = await repo.append(
+                "PRE_FLIGHT_PASSED" if allowed else "PRE_FLIGHT_FAILED",
+                incident_id=incident_id,
+                auto_repair_id=str(item.get("auto_repair_id") or "") or None,
+                actor="adr100_remediation_service",
+                action_detail=f"adr100_remediation_dry_run:{payload.get('mode')}"[:200],
+                success=allowed,
+                context=context,
+            )
+            if record is not None:
+                history["alert_operation_id"] = getattr(record, "id", None)
+        except Exception as exc:
+            logger.warning(
+                "adr100_remediation_alert_operation_history_failed",
+                incident_id=incident_id,
+                error=str(exc),
+            )
+
+        try:
+            timeline = self._timeline_service
+            if timeline is None:
+                from src.services.approval_db import get_timeline_service
+
+                timeline = get_timeline_service()
+            event = await timeline.add_event(
+                event_type="verifier",
+                status=_timeline_status(payload),
+                title="ADR-100 remediation dry-run",
+                description=_history_description(context),
+                actor="adr100_remediation_service",
+                actor_role=str(payload.get("mode") or "dry_run"),
+                incident_id=incident_id,
+            )
+            if event:
+                history["timeline_event_id"] = event.get("id")
+        except Exception as exc:
+            logger.warning(
+                "adr100_remediation_timeline_history_failed",
+                incident_id=incident_id,
+                error=str(exc),
+            )
+
+        history["recorded"] = bool(
+            history.get("alert_operation_id") or history.get("timeline_event_id")
+        )
+        return history
+
+
+def _select_mode(item: dict[str, Any], requested: RemediationMode) -> Literal["reverify", "replay"]:
+    if requested in ("reverify", "replay"):
+        return requested
+    if item.get("remediation_status") == "ready_for_reverify":
+        return "reverify"
+    if item.get("remediation_action") == "reverify_with_promql_template":
+        return "reverify"
+    return "replay"
+
+
+def _base_checks(item: dict[str, Any]) -> list[dict[str, Any]]:
+    status = str(item.get("remediation_status") or "unknown")
+    action = str(item.get("remediation_action") or "unknown")
+    return [
+        {
+            "name": "queue_item_ready",
+            "passed": status in _READY_STATUSES,
+            "detail": status,
+        },
+        {
+            "name": "read_only_guardrail",
+            "passed": action in {
+                "replay_with_supported_executor",
+                "reverify_with_promql_template",
+            },
+            "detail": action,
+        },
+        {
+            "name": "no_state_mutation",
+            "passed": True,
+            "detail": "dry_run_does_not_update_incident_or_auto_repair_rows",
+        },
+    ]
+
+
+def _plan_for_item(item: dict[str, Any], mode: str) -> dict[str, Any]:
+    if mode == "reverify":
+        return {
+            "step": "collect_current_state_and_assess",
+            "agent_id": "post_execution_verifier",
+            "required_scope": "read",
+            "writes": [],
+        }
+    return {
+        "step": "validate_supported_executor_route_then_collect_current_state",
+        "agent_id": "auto_repair_executor",
+        "required_scope": "read",
+        "writes": [],
+        "target_action": item.get("remediation_action"),
+    }
+
+
+def _dry_run_blocked_payload(
+    item: dict[str, Any],
+    mode: str,
+    checks: list[dict[str, Any]],
+) -> dict[str, Any]:
+    return {
+        "schema_version": "adr100_remediation_dry_run_v1",
+        "work_item_id": item.get("work_item_id"),
+        "incident_id": item.get("incident_id"),
+        "auto_repair_id": item.get("auto_repair_id"),
+        "mode": mode,
+        "allowed": False,
+        "executed": False,
+        "safety_level": "read_only",
+        "writes_incident_state": False,
+        "writes_auto_repair_result": False,
+        "checks": checks,
+        "verification_result_preview": "blocked",
+        "post_state_summary": {},
+    }
+
+
+def _dry_run_result_payload(
+    *,
+    item: dict[str, Any],
+    mode: str,
+    checks: list[dict[str, Any]],
+    post_state: dict[str, Any],
+    verification_result_preview: str,
+    extra: dict[str, Any],
+) -> dict[str, Any]:
+    return {
+        "schema_version": "adr100_remediation_dry_run_v1",
+        "work_item_id": item.get("work_item_id"),
+        "incident_id": item.get("incident_id"),
+        "auto_repair_id": item.get("auto_repair_id"),
+        "mode": mode,
+        "allowed": all(check["passed"] for check in checks),
+        "executed": True,
+        "safety_level": "read_only",
+        "writes_incident_state": False,
+        "writes_auto_repair_result": False,
+        "checks": checks,
+        "verification_result_preview": verification_result_preview,
+        "post_state_summary": _summarize_post_state(post_state),
+        **extra,
+    }
+
+
+def _summarize_post_state(post_state: dict[str, Any]) -> dict[str, Any]:
+    keys = sorted(post_state.keys())
+    return {
+        "tool_count": len(keys),
+        "tools": keys[:8],
+        "has_state": bool(post_state),
+    }
+
+
+def _history_context(item: dict[str, Any], payload: dict[str, Any]) -> dict[str, Any]:
+    return {
+        "schema_version": "adr100_remediation_dry_run_history_v1",
+        "work_item_id": item.get("work_item_id"),
+        "auto_repair_id": item.get("auto_repair_id"),
+        "playbook_id": item.get("playbook_id"),
+        "alertname": item.get("alertname"),
+        "mode": payload.get("mode"),
+        "allowed": payload.get("allowed"),
+        "executed": payload.get("executed"),
+        "safety_level": payload.get("safety_level"),
+        "writes_incident_state": payload.get("writes_incident_state"),
+        "writes_auto_repair_result": payload.get("writes_auto_repair_result"),
+        "verification_result_preview": payload.get("verification_result_preview"),
+        "post_state_summary": payload.get("post_state_summary"),
+        "mcp_route": payload.get("mcp_route"),
+        "checks": payload.get("checks"),
+    }
+
+
+def _timeline_status(payload: dict[str, Any]) -> str:
+    if not payload.get("allowed"):
+        return "warning"
+    if payload.get("verification_result_preview") == "success":
+        return "success"
+    return "warning"
+
+
+def _history_description(context: dict[str, Any]) -> str:
+    tool_count = (context.get("post_state_summary") or {}).get("tool_count", 0)
+    route = context.get("mcp_route") or {}
+    agent = route.get("agent_id") or "unknown_agent"
+    tool = route.get("tool_name") or "current_state"
+    return (
+        f"mode={context.get('mode')} "
+        f"preview={context.get('verification_result_preview')} "
+        f"tools={tool_count} route={agent}/{tool} "
+        f"writes_incident={context.get('writes_incident_state')} "
+        f"writes_auto_repair={context.get('writes_auto_repair_result')}"
+    )[:500]
+
+
+def _record_created_at(record: Any) -> str:
+    value = getattr(record, "created_at", None)
+    if hasattr(value, "isoformat"):
+        return value.isoformat()
+    return str(value or "")
+
+
+def _history_item(record: Any, context: dict[str, Any]) -> dict[str, Any]:
+    route = context.get("mcp_route") or {}
+    post_state = context.get("post_state_summary") or {}
+    return {
+        "id": str(getattr(record, "id", "")),
+        "incident_id": getattr(record, "incident_id", None),
+        "auto_repair_id": getattr(record, "auto_repair_id", None)
+        or context.get("auto_repair_id"),
+        "event_type": str(getattr(record, "event_type", "")),
+        "actor": getattr(record, "actor", None),
+        "success": getattr(record, "success", None),
+        "created_at": _record_created_at(record),
+        "work_item_id": context.get("work_item_id"),
+        "playbook_id": context.get("playbook_id"),
+        "alertname": context.get("alertname"),
+        "mode": context.get("mode"),
+        "allowed": context.get("allowed"),
+        "executed": context.get("executed"),
+        "safety_level": context.get("safety_level"),
+        "verification_result_preview": context.get("verification_result_preview"),
+        "tool_count": post_state.get("tool_count", 0),
+        "tools": post_state.get("tools") or [],
+        "agent_id": route.get("agent_id"),
+        "tool_name": route.get("tool_name") or "current_state",
+        "required_scope": route.get("required_scope"),
+        "writes_incident_state": context.get("writes_incident_state"),
+        "writes_auto_repair_result": context.get("writes_auto_repair_result"),
+        "checks": context.get("checks") or [],
+    }
+
+
+def _summarize_history_by_work_item(items: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    summary: dict[str, dict[str, Any]] = {}
+    for item in items:
+        key = str(item.get("work_item_id") or item.get("incident_id") or item.get("id"))
+        if key not in summary:
+            summary[key] = {
+                "work_item_id": item.get("work_item_id"),
+                "incident_id": item.get("incident_id"),
+                "count": 0,
+                "latest_at": item.get("created_at"),
+                "latest_event_type": item.get("event_type"),
+                "latest_success": item.get("success"),
+                "latest_preview": item.get("verification_result_preview"),
+                "latest_mode": item.get("mode"),
+                "latest_agent_id": item.get("agent_id"),
+                "latest_tool_name": item.get("tool_name"),
+                "required_scope": item.get("required_scope"),
+            }
+        summary[key]["count"] += 1
+    return list(summary.values())
+
+
+def _diagnostic_command_for_incident(incident: Incident) -> str:
+    labels = _labels_for_incident(incident)
+    host = str(labels.get("host") or labels.get("instance") or "{host}")
+    container = str(labels.get("container_name") or labels.get("container") or "")
+    if container:
+        return f"ssh {host} 'uptime; docker stats --no-stream {container}'"
+    return f"ssh {host} 'uptime; docker stats --no-stream'"
+
+
+def _promql_for_incident(incident: Incident) -> str:
+    labels = _labels_for_incident(incident)
+    alertname = ""
+    if incident.signals:
+        signal = incident.signals[0]
+        alertname = labels.get("alertname") or getattr(signal, "alert_name", "")
+    return _build_prometheus_query(alertname, labels)
+
+
+def _labels_for_incident(incident: Incident) -> dict[str, Any]:
+    if incident.signals:
+        return incident.signals[0].labels or {}
+    return {}
+
+
+_service: Adr100RemediationService | None = None
+
+
+def get_adr100_remediation_service() -> Adr100RemediationService:
+    """Return singleton ADR-100 remediation service."""
+
+    global _service
+    if _service is None:
+        _service = Adr100RemediationService()
+    return _service
+
+
+def set_adr100_remediation_service(service: Adr100RemediationService | None) -> None:
+    """Inject ADR-100 remediation service for tests."""
+
+    global _service
+    _service = service

apps/api/src/services/adr100_slo_metrics_service.py

@@ -0,0 +1,354 @@
+"""
+ADR-100 SLO metrics emitter.
+
+Prometheus recording rules for the AI flywheel SLOs expect a small set of
+counter-like metrics. The source of truth already lives in PostgreSQL, so this
+read-side emitter exposes DB totals on /metrics without changing runtime write
+paths or introducing another state store.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+from time import time
+
+from sqlalchemy import text
+
+from src.db.base import get_db_context
+
+
+@dataclass(frozen=True)
+class AutomationOperationSample:
+    outcome: str
+    operation_type: str
+    count: int
+
+
+@dataclass(frozen=True)
+class VerificationSample:
+    outcome: str
+    count: int
+
+
+@dataclass(frozen=True)
+class Adr100SloMetricsSnapshot:
+    automation_operations: list[AutomationOperationSample] = field(default_factory=list)
+    automation_operations_24h: list[AutomationOperationSample] = field(default_factory=list)
+    post_execution_verifications: list[VerificationSample] = field(default_factory=list)
+    post_execution_verifications_24h: list[VerificationSample] = field(default_factory=list)
+    knowledge_entries_total: int = 0
+    knowledge_entries_created_24h: int = 0
+    high_confidence_total: int = 0
+    high_confidence_success_total: int = 0
+    emitted_at: float = field(default_factory=time)
+
+
+class Adr100SloMetricsService:
+    """Build ADR-100 Prometheus samples from production DB state."""
+
+    async def to_prometheus_lines(self) -> str:
+        snapshot = await self.fetch_snapshot()
+        return render_adr100_slo_metrics(snapshot)
+
+    async def fetch_snapshot(self) -> Adr100SloMetricsSnapshot:
+        async with get_db_context() as db:
+            automation_rows = (
+                await db.execute(text(_AUTOMATION_OPERATION_SQL))
+            ).fetchall()
+            automation_24h_rows = (
+                await db.execute(text(_AUTOMATION_OPERATION_24H_SQL))
+            ).fetchall()
+            verification_rows = (
+                await db.execute(text(_POST_EXECUTION_VERIFICATION_SQL))
+            ).fetchall()
+            verification_24h_rows = (
+                await db.execute(text(_POST_EXECUTION_VERIFICATION_24H_SQL))
+            ).fetchall()
+            knowledge_total = int(
+                (await db.execute(text("SELECT count(*) FROM knowledge_entries"))).scalar()
+                or 0
+            )
+            knowledge_created_24h = int(
+                (
+                    await db.execute(
+                        text(
+                            """
+                            SELECT count(*)
+                            FROM knowledge_entries
+                            WHERE created_at >= NOW() - INTERVAL '24 hours'
+                            """
+                        )
+                    )
+                ).scalar()
+                or 0
+            )
+            confidence_row = (
+                await db.execute(text(_HIGH_CONFIDENCE_APPROVAL_SQL))
+            ).one()
+
+        return Adr100SloMetricsSnapshot(
+            automation_operations=[
+                AutomationOperationSample(
+                    outcome=str(row.outcome),
+                    operation_type=str(row.operation_type),
+                    count=int(row.count or 0),
+                )
+                for row in automation_rows
+            ],
+            automation_operations_24h=[
+                AutomationOperationSample(
+                    outcome=str(row.outcome),
+                    operation_type=str(row.operation_type),
+                    count=int(row.count or 0),
+                )
+                for row in automation_24h_rows
+            ],
+            post_execution_verifications=[
+                VerificationSample(
+                    outcome=str(row.outcome),
+                    count=int(row.count or 0),
+                )
+                for row in verification_rows
+            ],
+            post_execution_verifications_24h=[
+                VerificationSample(
+                    outcome=str(row.outcome),
+                    count=int(row.count or 0),
+                )
+                for row in verification_24h_rows
+            ],
+            knowledge_entries_total=knowledge_total,
+            knowledge_entries_created_24h=knowledge_created_24h,
+            high_confidence_total=int(confidence_row.high_confidence_total or 0),
+            high_confidence_success_total=int(
+                confidence_row.high_confidence_success_total or 0
+            ),
+        )
+
+
+def render_adr100_slo_metrics(snapshot: Adr100SloMetricsSnapshot) -> str:
+    """Render ADR-100 SLO metrics in Prometheus text exposition format."""
+    lines: list[str] = [
+        "",
+        "# HELP automation_operation_log_total DB-derived AI automation operation count for ADR-100 SLOs",
+        "# TYPE automation_operation_log_total counter",
+    ]
+    if snapshot.automation_operations:
+        for sample in snapshot.automation_operations:
+            lines.append(
+                "automation_operation_log_total"
+                f'{{outcome="{_escape_label(sample.outcome)}",'
+                f'operation_type="{_escape_label(sample.operation_type)}"}} '
+                f"{sample.count}"
+            )
+    else:
+        lines.append(
+            'automation_operation_log_total{outcome="none",operation_type="none"} 0'
+        )
+
+    lines.extend([
+        "# HELP automation_operation_created_24h DB-derived AI automation operation count created in the last 24 hours for ADR-100 SLO dashboards",
+        "# TYPE automation_operation_created_24h gauge",
+    ])
+    if snapshot.automation_operations_24h:
+        for sample in snapshot.automation_operations_24h:
+            lines.append(
+                "automation_operation_created_24h"
+                f'{{outcome="{_escape_label(sample.outcome)}",'
+                f'operation_type="{_escape_label(sample.operation_type)}"}} '
+                f"{sample.count}"
+            )
+    else:
+        lines.append(
+            'automation_operation_created_24h{outcome="none",operation_type="none"} 0'
+        )
+
+    lines.extend([
+        "# HELP post_execution_verification_total DB-derived post execution verification result count for ADR-100 SLOs",
+        "# TYPE post_execution_verification_total counter",
+    ])
+    if snapshot.post_execution_verifications:
+        for sample in snapshot.post_execution_verifications:
+            lines.append(
+                "post_execution_verification_total"
+                f'{{outcome="{_escape_label(sample.outcome)}"}} {sample.count}'
+            )
+    else:
+        lines.append('post_execution_verification_total{outcome="none"} 0')
+
+    lines.extend([
+        "# HELP post_execution_verification_created_24h DB-derived post execution verification result count created in the last 24 hours for ADR-100 SLO dashboards",
+        "# TYPE post_execution_verification_created_24h gauge",
+    ])
+    if snapshot.post_execution_verifications_24h:
+        for sample in snapshot.post_execution_verifications_24h:
+            lines.append(
+                "post_execution_verification_created_24h"
+                f'{{outcome="{_escape_label(sample.outcome)}"}} {sample.count}'
+            )
+    else:
+        lines.append('post_execution_verification_created_24h{outcome="none"} 0')
+
+    lines.extend([
+        "# HELP knowledge_entries_total DB-derived knowledge entry count for ADR-100 SLOs",
+        "# TYPE knowledge_entries_total counter",
+        f"knowledge_entries_total {snapshot.knowledge_entries_total}",
+        "# HELP knowledge_entries_created_24h DB-derived knowledge entries created in the last 24 hours for ADR-100 SLOs",
+        "# TYPE knowledge_entries_created_24h gauge",
+        f"knowledge_entries_created_24h {snapshot.knowledge_entries_created_24h}",
+        "# HELP approval_records_high_confidence_total DB-derived high confidence approval decisions for ADR-100 SLOs",
+        "# TYPE approval_records_high_confidence_total counter",
+        f"approval_records_high_confidence_total {snapshot.high_confidence_total}",
+        "# HELP approval_records_high_confidence_success_total DB-derived high confidence approval decisions with successful verification for ADR-100 SLOs",
+        "# TYPE approval_records_high_confidence_success_total counter",
+        (
+            "approval_records_high_confidence_success_total "
+            f"{snapshot.high_confidence_success_total}"
+        ),
+        "# HELP adr100_slo_emitter_last_success_timestamp Last successful ADR-100 DB metrics emission timestamp",
+        "# TYPE adr100_slo_emitter_last_success_timestamp gauge",
+        f"adr100_slo_emitter_last_success_timestamp {snapshot.emitted_at:.0f}",
+        "",
+    ])
+    return "\n".join(lines)
+
+
+def _escape_label(value: str) -> str:
+    return value.replace("\\", "\\\\").replace("\n", "\\n").replace('"', '\\"')
+
+
+_AUTOMATION_OPERATION_SQL = """
+    WITH automation_scope AS (
+        SELECT
+            CASE
+                WHEN status <> 'success' THEN status
+                WHEN actor = 'approval_execution'
+                     AND COALESCE(input->>'requested_by', '') NOT ILIKE 'auto%%'
+                    THEN 'human_required'
+                ELSE 'auto_executed'
+            END AS outcome,
+            operation_type
+        FROM automation_operation_log
+        WHERE operation_type IN (
+            'playbook_executed',
+            'remediation_executed',
+            'remediation_verified',
+            'remediation_rolled_back',
+            'self_correction_attempted'
+        )
+        UNION ALL
+        SELECT
+            CASE WHEN success THEN 'auto_executed' ELSE 'failed' END AS outcome,
+            'auto_repair_executed' AS operation_type
+        FROM auto_repair_executions
+    )
+    SELECT
+        outcome,
+        operation_type,
+        count(*) AS count
+    FROM automation_scope
+    GROUP BY outcome, operation_type
+    ORDER BY outcome, operation_type
+"""
+
+
+_AUTOMATION_OPERATION_24H_SQL = """
+    WITH automation_scope AS (
+        SELECT
+            CASE
+                WHEN status <> 'success' THEN status
+                WHEN actor = 'approval_execution'
+                     AND COALESCE(input->>'requested_by', '') NOT ILIKE 'auto%%'
+                    THEN 'human_required'
+                ELSE 'auto_executed'
+            END AS outcome,
+            operation_type
+        FROM automation_operation_log
+        WHERE operation_type IN (
+            'playbook_executed',
+            'remediation_executed',
+            'remediation_verified',
+            'remediation_rolled_back',
+            'self_correction_attempted'
+        )
+          AND created_at >= NOW() - INTERVAL '24 hours'
+        UNION ALL
+        SELECT
+            CASE WHEN success THEN 'auto_executed' ELSE 'failed' END AS outcome,
+            'auto_repair_executed' AS operation_type
+        FROM auto_repair_executions
+        WHERE created_at >= NOW() - INTERVAL '24 hours'
+    )
+    SELECT
+        outcome,
+        operation_type,
+        count(*) AS count
+    FROM automation_scope
+    GROUP BY outcome, operation_type
+    ORDER BY outcome, operation_type
+"""
+
+
+_POST_EXECUTION_VERIFICATION_SQL = """
+    SELECT verification_result AS outcome, count(*) AS count
+    FROM incident_evidence
+    WHERE verification_result IS NOT NULL
+    GROUP BY verification_result
+    ORDER BY verification_result
+"""
+
+
+_POST_EXECUTION_VERIFICATION_24H_SQL = """
+    SELECT verification_result AS outcome, count(*) AS count
+    FROM incident_evidence
+    WHERE verification_result IS NOT NULL
+      AND collected_at >= NOW() - INTERVAL '24 hours'
+    GROUP BY verification_result
+    ORDER BY verification_result
+"""
+
+
+_HIGH_CONFIDENCE_APPROVAL_SQL = """
+    WITH approval_confidence AS (
+        SELECT
+            id,
+            incident_id,
+            COALESCE(
+                CASE
+                    WHEN extra_metadata->>'confidence_score' ~ '^[0-9]+(\\.[0-9]+)?$'
+                        THEN (extra_metadata->>'confidence_score')::numeric
+                    ELSE NULL
+                END,
+                CASE
+                    WHEN extra_metadata->>'confidence' ~ '^[0-9]+(\\.[0-9]+)?$'
+                        THEN (extra_metadata->>'confidence')::numeric
+                    ELSE NULL
+                END,
+                composite_score,
+                0
+            ) AS confidence
+        FROM approval_records
+    )
+    SELECT
+        count(*) FILTER (WHERE confidence >= 0.8) AS high_confidence_total,
+        count(*) FILTER (
+            WHERE confidence >= 0.8
+              AND EXISTS (
+                  SELECT 1
+                  FROM incident_evidence ev
+                  WHERE ev.incident_id = approval_confidence.incident_id
+                    AND ev.verification_result = 'success'
+              )
+        ) AS high_confidence_success_total
+    FROM approval_confidence
+"""
+
+
+_adr100_slo_metrics_service: Adr100SloMetricsService | None = None
+
+
+def get_adr100_slo_metrics_service() -> Adr100SloMetricsService:
+    global _adr100_slo_metrics_service
+    if _adr100_slo_metrics_service is None:
+        _adr100_slo_metrics_service = Adr100SloMetricsService()
+    return _adr100_slo_metrics_service

apps/api/src/services/adr100_slo_status_service.py

@@ -0,0 +1,743 @@
+"""
+Read-only ADR-100 SLO status snapshot.
+
+GovernanceAgent.check_slo_compliance() can emit governance alerts when an SLO is
+violated. This service is intentionally read-only so dashboards can show the
+same Prometheus-backed state without producing Telegram/DB side effects.
+"""
+
+from __future__ import annotations
+
+import math
+from dataclasses import dataclass
+from typing import Any
+
+import httpx
+import structlog
+from sqlalchemy import text
+
+from src.core.config import settings
+from src.db.base import get_db_context
+from src.utils.timezone import now_taipei_iso
+
+logger = structlog.get_logger(__name__)
+
+
+@dataclass(frozen=True)
+class Adr100SloDefinition:
+    name: str
+    query: str
+    target: float
+    hard_red_line: float
+    direction: str
+    unit: str
+    window: str
+    denominator_query: str | None = None
+    denominator_window_seconds: int = 0
+    minimum_events: float = 1.0
+
+
+ADR100_SLO_DEFINITIONS: tuple[Adr100SloDefinition, ...] = (
+    Adr100SloDefinition(
+        name="autonomy_rate",
+        query="sli:autonomy_rate:5m",
+        target=0.80,
+        hard_red_line=0.70,
+        direction="above",
+        unit="percent",
+        window="5m",
+        denominator_query="sum(rate(automation_operation_log_total[5m]))",
+        denominator_window_seconds=300,
+    ),
+    Adr100SloDefinition(
+        name="decision_accuracy",
+        query="sli:decision_accuracy:5m",
+        target=0.90,
+        hard_red_line=0.85,
+        direction="above",
+        unit="percent",
+        window="5m",
+        denominator_query='sum(rate(automation_operation_log_total{outcome="auto_executed"}[5m]))',
+        denominator_window_seconds=300,
+    ),
+    Adr100SloDefinition(
+        name="confidence_calibration",
+        query="sli:confidence_calibration:1h",
+        target=0.80,
+        hard_red_line=0.70,
+        direction="above",
+        unit="percent",
+        window="1h",
+        denominator_query="sum(rate(approval_records_high_confidence_total[1h]))",
+        denominator_window_seconds=3600,
+    ),
+    Adr100SloDefinition(
+        name="km_growth_rate",
+        query="max(knowledge_entries_created_24h) or max(sli:km_growth_rate:24h)",
+        target=20.0,
+        hard_red_line=5.0,
+        direction="above",
+        unit="count",
+        window="24h",
+    ),
+)
+
+
+class Adr100SloStatusService:
+    """Fetch ADR-100 SLO status from Prometheus without writing governance events."""
+
+    async def fetch_report(self) -> dict[str, Any]:
+        prom_url = getattr(
+            settings,
+            "PROMETHEUS_URL",
+            "http://prometheus.observability.svc:9090",
+        ).rstrip("/")
+        metrics: list[dict[str, Any]] = []
+
+        async with httpx.AsyncClient(timeout=5.0) as client:
+            for definition in ADR100_SLO_DEFINITIONS:
+                metrics.append(await self._fetch_metric(client, prom_url, definition))
+
+        evaluable = [metric for metric in metrics if metric.get("evaluable")]
+        ok_count = sum(1 for metric in evaluable if metric.get("status") == "ok")
+        overall_compliance = (ok_count / len(evaluable)) if evaluable else None
+        verification_coverage = await self._fetch_verification_coverage()
+        overall_status = _overall_status(metrics, evaluable, verification_coverage)
+
+        return {
+            "schema_version": "adr100_slo_status_v1",
+            "source": "prometheus+postgresql",
+            "evaluated_at": now_taipei_iso(),
+            "overall_status": overall_status,
+            "overall_compliance": overall_compliance,
+            "evaluable_count": len(evaluable),
+            "metric_count": len(metrics),
+            "metrics": metrics,
+            "verification_coverage": verification_coverage,
+        }
+
+    async def _fetch_metric(
+        self,
+        client: httpx.AsyncClient,
+        prom_url: str,
+        definition: Adr100SloDefinition,
+    ) -> dict[str, Any]:
+        denominator_value: float | None = None
+        sample_count: float | None = None
+
+        if definition.denominator_query:
+            denominator_result = await _query_prometheus_value(
+                client,
+                prom_url,
+                definition.denominator_query,
+            )
+            if denominator_result["status"] != "ok":
+                return _metric_payload(
+                    definition,
+                    value=None,
+                    status="no_data",
+                    reason=denominator_result["reason"],
+                    denominator_value=None,
+                    sample_count=None,
+                )
+
+            denominator_value = float(denominator_result["value"])
+            sample_count = denominator_value * definition.denominator_window_seconds
+            if sample_count < definition.minimum_events:
+                return _metric_payload(
+                    definition,
+                    value=None,
+                    status="skipped_low_volume",
+                    reason="denominator_below_minimum_events",
+                    denominator_value=denominator_value,
+                    sample_count=sample_count,
+                )
+
+        value_result = await _query_prometheus_value(client, prom_url, definition.query)
+        if value_result["status"] != "ok":
+            status = (
+                "skipped_low_volume"
+                if value_result["reason"] == "prometheus_nan_or_inf"
+                else "no_data"
+            )
+            return _metric_payload(
+                definition,
+                value=None,
+                status=status,
+                reason=value_result["reason"],
+                denominator_value=denominator_value,
+                sample_count=sample_count,
+            )
+
+        value = float(value_result["value"])
+        status = _classify_status(value, definition)
+        return _metric_payload(
+            definition,
+            value=value,
+            status=status,
+            reason=None,
+            denominator_value=denominator_value,
+            sample_count=sample_count if sample_count is not None else value,
+        )
+
+    async def _fetch_verification_coverage(self) -> dict[str, Any]:
+        """Summarize whether recent auto-repair executions have verifier evidence."""
+        try:
+            async with get_db_context() as db:
+                summary_row = (
+                    await db.execute(text(_VERIFICATION_COVERAGE_SQL))
+                ).mappings().one()
+                recent_rows = (
+                    await db.execute(text(_VERIFICATION_COVERAGE_RECENT_SQL))
+                ).mappings().all()
+                recent_non_success_rows = (
+                    await db.execute(text(_VERIFICATION_COVERAGE_NON_SUCCESS_SQL))
+                ).mappings().all()
+        except Exception as exc:
+            logger.warning("adr100_verification_coverage_query_error", error=str(exc))
+            return {
+                "schema_version": "adr100_verification_coverage_v1",
+                "source": "postgresql",
+                "window": "24h",
+                "status": "error",
+                "reason": "postgresql_query_error",
+                "evaluable": False,
+                "total_auto": 0,
+                "successful_auto": 0,
+                "verified_auto": 0,
+                "verified_success": 0,
+                "verified_non_success": 0,
+                "unverified_auto": 0,
+                "coverage_rate": None,
+                "verification_success_rate": None,
+                "last_auto_at": None,
+                "last_verified_auto_at": None,
+                "last_verification_evidence_at": None,
+                "latest_auto_age_seconds": None,
+                "last_verified_auto_age_seconds": None,
+                "recent_unverified": [],
+                "recent_non_success": [],
+                "non_success_breakdown": {
+                    "by_verification_result": [],
+                    "by_failure_class": [],
+                },
+                "remediation_queue": _remediation_queue_payload([]),
+            }
+
+        return _build_verification_coverage_payload(
+            summary_row,
+            recent_rows,
+            recent_non_success_rows,
+        )
+
+
+_VERIFICATION_COVERAGE_SQL = """
+    WITH recent_auto AS (
+        SELECT id, incident_id, success, created_at
+        FROM auto_repair_executions
+        WHERE created_at >= NOW() - INTERVAL '24 hours'
+    ),
+    per_auto AS (
+        SELECT
+            are.id,
+            are.incident_id,
+            are.success,
+            are.created_at,
+            latest.verification_result,
+            latest.collected_at AS verification_collected_at,
+            latest.self_healing_score
+        FROM recent_auto are
+        LEFT JOIN LATERAL (
+            SELECT ev.verification_result, ev.collected_at, ev.self_healing_score
+            FROM incident_evidence ev
+            WHERE ev.incident_id = are.incident_id
+              AND ev.verification_result IS NOT NULL
+            ORDER BY ev.collected_at DESC
+            LIMIT 1
+        ) latest ON TRUE
+    )
+    SELECT
+        count(*)::int AS total_auto,
+        count(*) FILTER (WHERE success)::int AS successful_auto,
+        count(*) FILTER (WHERE verification_result IS NOT NULL)::int AS verified_auto,
+        count(*) FILTER (WHERE verification_result = 'success')::int AS verified_success,
+        count(*) FILTER (WHERE verification_result IN ('degraded','failed','timeout'))::int AS verified_non_success,
+        count(*) FILTER (WHERE verification_result IS NULL)::int AS unverified_auto,
+        max(created_at) AS last_auto_at,
+        max(created_at) FILTER (WHERE verification_result IS NOT NULL) AS last_verified_auto_at,
+        max(verification_collected_at) AS last_verification_evidence_at,
+        EXTRACT(EPOCH FROM (NOW() - max(created_at)))::int AS latest_auto_age_seconds,
+        EXTRACT(EPOCH FROM (NOW() - (max(created_at) FILTER (WHERE verification_result IS NOT NULL))))::int
+            AS last_verified_auto_age_seconds
+    FROM per_auto
+"""
+
+
+_VERIFICATION_COVERAGE_RECENT_SQL = """
+    WITH recent_auto AS (
+        SELECT id, incident_id, success, created_at
+        FROM auto_repair_executions
+        WHERE created_at >= NOW() - INTERVAL '24 hours'
+    ),
+    per_auto AS (
+        SELECT
+            are.id,
+            are.incident_id,
+            are.success,
+            are.created_at,
+            latest.verification_result
+        FROM recent_auto are
+        LEFT JOIN LATERAL (
+            SELECT ev.verification_result
+            FROM incident_evidence ev
+            WHERE ev.incident_id = are.incident_id
+              AND ev.verification_result IS NOT NULL
+            ORDER BY ev.collected_at DESC
+            LIMIT 1
+        ) latest ON TRUE
+    )
+    SELECT id, incident_id, success, created_at
+    FROM per_auto
+    WHERE verification_result IS NULL
+    ORDER BY created_at DESC
+    LIMIT 5
+"""
+
+
+_VERIFICATION_COVERAGE_NON_SUCCESS_SQL = """
+    WITH recent_auto AS (
+        SELECT
+            id,
+            incident_id,
+            success,
+            playbook_id,
+            playbook_name,
+            triggered_by,
+            risk_level,
+            error_message,
+            created_at
+        FROM auto_repair_executions
+        WHERE created_at >= NOW() - INTERVAL '24 hours'
+    ),
+    per_auto AS (
+        SELECT
+            are.id AS auto_repair_id,
+            are.incident_id,
+            are.success AS auto_success,
+            are.playbook_id,
+            are.playbook_name,
+            are.triggered_by,
+            are.risk_level,
+            left(coalesce(are.error_message, ''), 240) AS auto_error,
+            are.created_at AS auto_created_at,
+            latest.verification_result,
+            latest.collected_at AS verification_collected_at,
+            left(coalesce(latest.post_execution_state::text, ''), 700) AS post_state_text,
+            left(coalesce(latest.evidence_summary, ''), 300) AS evidence_summary
+        FROM recent_auto are
+        LEFT JOIN LATERAL (
+            SELECT
+                ev.verification_result,
+                ev.collected_at,
+                ev.post_execution_state,
+                ev.evidence_summary
+            FROM incident_evidence ev
+            WHERE ev.incident_id = are.incident_id
+              AND ev.verification_result IS NOT NULL
+            ORDER BY ev.collected_at DESC
+            LIMIT 1
+        ) latest ON TRUE
+    )
+    SELECT
+        p.*,
+        i.status::text AS incident_status,
+        i.severity::text AS incident_severity,
+        i.alert_category,
+        i.alertname
+    FROM per_auto p
+    LEFT JOIN incidents i ON i.incident_id = p.incident_id
+    WHERE p.verification_result IS NOT NULL
+      AND p.verification_result <> 'success'
+    ORDER BY p.auto_created_at DESC
+    LIMIT 8
+"""
+
+
+async def _query_prometheus_value(
+    client: httpx.AsyncClient,
+    prom_url: str,
+    query: str,
+) -> dict[str, Any]:
+    try:
+        response = await client.get(
+            f"{prom_url}/api/v1/query",
+            params={"query": query},
+        )
+        data = response.json()
+        if data.get("status") != "success":
+            return {"status": "error", "reason": "prometheus_query_failed"}
+
+        results = data.get("data", {}).get("result", [])
+        if not results:
+            return {
+                "status": "no_data",
+                "reason": "prometheus_empty_result_metric_not_emitted",
+            }
+
+        raw_value = results[0]["value"][1]
+        value = float(raw_value)
+        if not math.isfinite(value):
+            return {
+                "status": "skipped",
+                "reason": "prometheus_nan_or_inf",
+                "raw_value": raw_value,
+            }
+        return {"status": "ok", "value": value}
+    except Exception as exc:
+        logger.warning("adr100_slo_prometheus_query_error", query=query, error=str(exc))
+        return {"status": "error", "reason": "prometheus_query_error"}
+
+
+def _metric_payload(
+    definition: Adr100SloDefinition,
+    *,
+    value: float | None,
+    status: str,
+    reason: str | None,
+    denominator_value: float | None,
+    sample_count: float | None,
+) -> dict[str, Any]:
+    return {
+        "name": definition.name,
+        "query": definition.query,
+        "value": value,
+        "target": definition.target,
+        "hard_red_line": definition.hard_red_line,
+        "direction": definition.direction,
+        "unit": definition.unit,
+        "window": definition.window,
+        "status": status,
+        "evaluable": status in {"ok", "warning", "violated"},
+        "reason": reason,
+        "denominator_query": definition.denominator_query,
+        "denominator_value": denominator_value,
+        "sample_count": sample_count,
+    }
+
+
+def _classify_status(value: float, definition: Adr100SloDefinition) -> str:
+    if definition.direction == "above":
+        if value < definition.hard_red_line:
+            return "violated"
+        if value < definition.target:
+            return "warning"
+        return "ok"
+
+    if value > definition.hard_red_line:
+        return "violated"
+    if value > definition.target:
+        return "warning"
+    return "ok"
+
+
+def _build_verification_coverage_payload(
+    summary_row: Any,
+    recent_unverified_rows: Any,
+    recent_non_success_rows: Any = (),
+) -> dict[str, Any]:
+    row = dict(summary_row)
+    total_auto = int(row.get("total_auto") or 0)
+    verified_auto = int(row.get("verified_auto") or 0)
+    verified_success = int(row.get("verified_success") or 0)
+    verified_non_success = int(row.get("verified_non_success") or 0)
+    unverified_auto = int(row.get("unverified_auto") or 0)
+
+    if total_auto == 0:
+        status = "skipped_low_volume"
+        reason = "no_auto_repair_executions_24h"
+        evaluable = False
+    elif unverified_auto > 0:
+        status = "warning"
+        reason = "verification_backlog_present"
+        evaluable = True
+    elif verified_non_success > 0:
+        status = "warning"
+        reason = "non_success_verification_present"
+        evaluable = True
+    else:
+        status = "ok"
+        reason = None
+        evaluable = True
+
+    coverage_rate = (verified_auto / total_auto) if total_auto else None
+    verification_success_rate = (verified_success / verified_auto) if verified_auto else None
+    recent_non_success = [
+        _non_success_finding_payload(dict(raw))
+        for raw in recent_non_success_rows
+    ]
+    remediation_queue = _remediation_queue_payload(recent_non_success)
+
+    return {
+        "schema_version": "adr100_verification_coverage_v1",
+        "source": "postgresql",
+        "window": "24h",
+        "status": status,
+        "reason": reason,
+        "evaluable": evaluable,
+        "total_auto": total_auto,
+        "successful_auto": int(row.get("successful_auto") or 0),
+        "verified_auto": verified_auto,
+        "verified_success": verified_success,
+        "verified_non_success": verified_non_success,
+        "unverified_auto": unverified_auto,
+        "coverage_rate": coverage_rate,
+        "verification_success_rate": verification_success_rate,
+        "last_auto_at": _iso(row.get("last_auto_at")),
+        "last_verified_auto_at": _iso(row.get("last_verified_auto_at")),
+        "last_verification_evidence_at": _iso(row.get("last_verification_evidence_at")),
+        "latest_auto_age_seconds": _int_or_none(row.get("latest_auto_age_seconds")),
+        "last_verified_auto_age_seconds": _int_or_none(row.get("last_verified_auto_age_seconds")),
+        "recent_unverified": [
+            {
+                "id": str(item.get("id")),
+                "incident_id": str(item.get("incident_id")),
+                "success": bool(item.get("success")),
+                "created_at": _iso(item.get("created_at")),
+            }
+            for item in (dict(raw) for raw in recent_unverified_rows)
+        ],
+        "recent_non_success": recent_non_success,
+        "non_success_breakdown": {
+            "by_verification_result": _count_breakdown(
+                item["verification_result"] for item in recent_non_success
+            ),
+            "by_failure_class": _count_breakdown(
+                item["failure_class"] for item in recent_non_success
+            ),
+            "by_remediation_status": _count_breakdown(
+                item["remediation_status"] for item in remediation_queue["items"]
+            ),
+        },
+        "remediation_queue": remediation_queue,
+    }
+
+
+def _non_success_finding_payload(row: dict[str, Any]) -> dict[str, Any]:
+    failure_class = _classify_non_success_failure(row)
+    remediation = _remediation_for_failure_class(failure_class)
+    return {
+        "auto_repair_id": str(row.get("auto_repair_id")),
+        "incident_id": str(row.get("incident_id")),
+        "incident_status": str(row.get("incident_status") or "unknown"),
+        "incident_severity": str(row.get("incident_severity") or "unknown"),
+        "alert_category": row.get("alert_category"),
+        "alertname": row.get("alertname"),
+        "auto_success": bool(row.get("auto_success")),
+        "playbook_id": row.get("playbook_id"),
+        "playbook_name": row.get("playbook_name"),
+        "triggered_by": row.get("triggered_by"),
+        "risk_level": row.get("risk_level"),
+        "verification_result": str(row.get("verification_result") or "unknown"),
+        "failure_class": failure_class,
+        "next_step": _next_step_for_failure_class(failure_class),
+        "remediation_status": remediation["status"],
+        "remediation_action": remediation["action"],
+        "remediation_owner": remediation["owner"],
+        "remediation_reason": remediation["reason"],
+        "auto_error_excerpt": _short_text(row.get("auto_error"), 180),
+        "evidence_excerpt": _short_text(row.get("evidence_summary"), 180),
+        "auto_created_at": _iso(row.get("auto_created_at")),
+        "verification_collected_at": _iso(row.get("verification_collected_at")),
+    }
+
+
+def _classify_non_success_failure(row: dict[str, Any]) -> str:
+    combined = " ".join(
+        str(row.get(key) or "")
+        for key in ("auto_error", "post_state_text", "evidence_summary")
+    ).lower()
+    if "unsupported scheme" in combined:
+        return "unsupported_action_scheme"
+    if "missing_query_parameter" in combined:
+        return "verifier_missing_promql"
+    if "empty_pod_name" in combined:
+        return "verifier_target_missing_pod"
+    if not bool(row.get("auto_success")):
+        return "auto_repair_execution_failed"
+
+    result = str(row.get("verification_result") or "").lower()
+    if result in {"failed", "timeout"}:
+        return f"verification_{result}"
+    return "verification_degraded"
+
+
+def _remediation_for_failure_class(failure_class: str) -> dict[str, str]:
+    """Map a non-success verification class to a read-only remediation work item.
+
+    This is dashboard triage metadata only. It does not auto-close incidents,
+    replay repairs, or approve write actions.
+    """
+    if failure_class == "unsupported_action_scheme":
+        return {
+            "status": "ready_for_replay",
+            "action": "replay_with_supported_executor",
+            "owner": "auto_repair_executor",
+            "reason": "executor_gateway_available_after_t23",
+        }
+    if failure_class == "verifier_missing_promql":
+        return {
+            "status": "ready_for_reverify",
+            "action": "reverify_with_promql_template",
+            "owner": "post_execution_verifier",
+            "reason": "promql_template_available_after_t23",
+        }
+    if failure_class == "verifier_target_missing_pod":
+        return {
+            "status": "needs_target_mapping",
+            "action": "map_target_and_reverify",
+            "owner": "post_execution_verifier",
+            "reason": "verifier_target_missing",
+        }
+    if failure_class == "auto_repair_execution_failed":
+        return {
+            "status": "needs_playbook_ticket",
+            "action": "create_playbook_ticket",
+            "owner": "solver_or_operator",
+            "reason": "execution_failed_after_route_normalization",
+        }
+    if failure_class in {"verification_failed", "verification_timeout"}:
+        return {
+            "status": "manual_review",
+            "action": "escalate_verification_failure",
+            "owner": "sre_operator",
+            "reason": "verifier_returned_hard_failure",
+        }
+    return {
+        "status": "manual_review",
+        "action": "inspect_degraded_evidence",
+        "owner": "sre_operator",
+        "reason": "degraded_evidence_requires_human_context",
+    }
+
+
+def _next_step_for_failure_class(failure_class: str) -> str:
+    if failure_class == "unsupported_action_scheme":
+        return "normalize_playbook_executor"
+    if failure_class == "verifier_missing_promql":
+        return "add_verifier_query_template"
+    if failure_class == "verifier_target_missing_pod":
+        return "map_verifier_target"
+    if failure_class == "auto_repair_execution_failed":
+        return "review_auto_repair_execution"
+    if failure_class in {"verification_failed", "verification_timeout"}:
+        return "escalate_verification_failure"
+    return "review_degraded_verification"
+
+
+def _remediation_queue_payload(recent_non_success: list[dict[str, Any]]) -> dict[str, Any]:
+    items: list[dict[str, Any]] = []
+    for item in recent_non_success:
+        items.append({
+            "work_item_id": (
+                f"verification:{item.get('incident_id')}:{item.get('auto_repair_id')}"
+            ),
+            "incident_id": item.get("incident_id"),
+            "auto_repair_id": item.get("auto_repair_id"),
+            "alertname": item.get("alertname"),
+            "playbook_id": item.get("playbook_id"),
+            "failure_class": item.get("failure_class"),
+            "verification_result": item.get("verification_result"),
+            "remediation_status": item.get("remediation_status"),
+            "remediation_action": item.get("remediation_action"),
+            "remediation_owner": item.get("remediation_owner"),
+            "remediation_reason": item.get("remediation_reason"),
+            "source": "adr100_verification_coverage",
+            "auto_created_at": item.get("auto_created_at"),
+            "verification_collected_at": item.get("verification_collected_at"),
+        })
+
+    ready_for_ai = sum(
+        1 for item in items
+        if item.get("remediation_status") in {"ready_for_replay", "ready_for_reverify"}
+    )
+    needs_human = sum(
+        1 for item in items
+        if item.get("remediation_status") in {
+            "needs_target_mapping",
+            "needs_playbook_ticket",
+            "manual_review",
+        }
+    )
+
+    return {
+        "schema_version": "adr100_remediation_queue_v1",
+        "source": "recent_non_success_read_model",
+        "total": len(items),
+        "ready_for_ai": ready_for_ai,
+        "needs_human": needs_human,
+        "items": items,
+        "by_status": _count_breakdown(
+            item.get("remediation_status") for item in items
+        ),
+        "by_action": _count_breakdown(
+            item.get("remediation_action") for item in items
+        ),
+    }
+
+
+def _count_breakdown(values: Any) -> list[dict[str, Any]]:
+    counts: dict[str, int] = {}
+    for value in values:
+        key = str(value or "unknown")
+        counts[key] = counts.get(key, 0) + 1
+    return [
+        {"name": name, "count": count}
+        for name, count in sorted(counts.items(), key=lambda item: (-item[1], item[0]))
+    ]
+
+
+def _short_text(value: Any, limit: int) -> str | None:
+    if value is None:
+        return None
+    text = " ".join(str(value).split())
+    if not text:
+        return None
+    return text[:limit]
+
+
+def _iso(value: Any) -> str | None:
+    return value.isoformat() if hasattr(value, "isoformat") else None
+
+
+def _int_or_none(value: Any) -> int | None:
+    return int(value) if value is not None else None
+
+
+def _overall_status(
+    metrics: list[dict[str, Any]],
+    evaluable: list[dict[str, Any]],
+    verification_coverage: dict[str, Any] | None = None,
+) -> str:
+    if any(metric.get("status") == "violated" for metric in metrics):
+        return "violated"
+    if verification_coverage and verification_coverage.get("status") in {"violated", "warning"}:
+        return str(verification_coverage["status"])
+    if any(metric.get("status") == "warning" for metric in metrics):
+        return "warning"
+    if evaluable and any(metric.get("status") == "skipped_low_volume" for metric in metrics):
+        return "partial"
+    if evaluable:
+        return "ok"
+    if any(metric.get("status") == "no_data" for metric in metrics):
+        return "no_data"
+    return "skipped_low_volume"
+
+
+_adr100_slo_status_service: Adr100SloStatusService | None = None
+
+
+def get_adr100_slo_status_service() -> Adr100SloStatusService:
+    global _adr100_slo_status_service
+    if _adr100_slo_status_service is None:
+        _adr100_slo_status_service = Adr100SloStatusService()
+    return _adr100_slo_status_service

apps/api/src/services/ai_router.py

@@ -842,14 +842,13 @@ class AIRouter:
             空 dict 代表無資料或查詢失敗（caller 應降級為忽略）。
         """
         try:
-            from src.db.base import get_session_factory
+            from src.db.base import get_db_context
             from src.repositories.aider_event_repository import AiderEventRepository
         except ImportError:
             return {}
 
         try:
-            sf = get_session_factory()
-            async with sf() as sess:
+            async with get_db_context() as sess:
                 repo_obj = AiderEventRepository(sess)
                 stats = await repo_obj.model_stats_since(days=days)
         except Exception:

apps/api/src/services/ai_slo_calculator.py

@@ -28,7 +28,7 @@ from datetime import timedelta
 import structlog
 from sqlalchemy import func, select, text
 
-from src.db.base import get_session_factory
+from src.db.base import get_db_context
 from src.db.models import AiGovernanceEvent, AutoRepairExecution, ApprovalRecord
 from src.utils.timezone import now_taipei
 
@@ -127,7 +127,7 @@ class AiSloCalculator:
         try:
             since = now_taipei() - timedelta(days=SLO_WINDOW_DAYS)
 
-            async with get_session_factory()() as session:
+            async with get_db_context() as session:
                 slo1 = await self._calc_auto_success_rate(session, since)
                 slo2 = await self._calc_human_override_rate(session, since)
                 slo3 = await self._calc_false_neg_rate(session, since)
@@ -210,7 +210,7 @@ class AiSloCalculator:
         只在 any_violated=True 時呼叫。不管舊違反是否解決。
         """
         try:
-            async with get_session_factory()() as session:
+            async with get_db_context() as session:
                 event = AiGovernanceEvent(
                     event_type="slo_violation",
                     details=report.to_dict(),

apps/api/src/services/alert_chain_metrics_service.py

@@ -0,0 +1,160 @@
+"""DB-backed Alert Chain metric refresh.
+
+`record_alert_chain_success()` updates Prometheus gauges in process memory. A
+deployment or Pod restart clears that memory, so the first scrape after deploy
+can report `awoooi_alert_chain_last_success_timestamp` as missing even though
+Alertmanager / Sentry / SigNoz evidence is already in DB.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from time import time
+from typing import Any, Mapping
+
+import structlog
+from sqlalchemy import text
+
+from src.core.metrics import ALERT_CHAIN_LAST_SUCCESS
+from src.db.base import get_db_context
+
+logger = structlog.get_logger(__name__)
+
+_REFRESH_INTERVAL_SECONDS = 15
+
+
+@dataclass(frozen=True)
+class AlertChainEvidenceSample:
+    source: str
+    last_success_timestamp: float
+    evidence_source: str
+
+
+class AlertChainMetricsService:
+    """Refresh process-local alert-chain gauges from durable DB evidence."""
+
+    def __init__(self) -> None:
+        self._last_refresh_at = 0.0
+        self._last_samples: list[AlertChainEvidenceSample] = []
+
+    async def refresh_last_success_gauge(
+        self,
+        *,
+        project_id: str = "awoooi",
+        force: bool = False,
+    ) -> list[AlertChainEvidenceSample]:
+        now_ts = time()
+        if (
+            not force
+            and self._last_samples
+            and now_ts - self._last_refresh_at < _REFRESH_INTERVAL_SECONDS
+        ):
+            apply_alert_chain_last_success_samples(self._last_samples)
+            return self._last_samples
+
+        samples = await self.fetch_latest_success_samples(project_id=project_id)
+        apply_alert_chain_last_success_samples(samples)
+        self._last_samples = samples
+        self._last_refresh_at = now_ts
+        return samples
+
+    async def fetch_latest_success_samples(
+        self,
+        *,
+        project_id: str = "awoooi",
+    ) -> list[AlertChainEvidenceSample]:
+        async with get_db_context(project_id) as db:
+            result = await db.execute(text(_ALERT_CHAIN_EVIDENCE_SQL), {"project_id": project_id})
+            return merge_alert_chain_evidence_rows(result.mappings().all())
+
+
+def apply_alert_chain_last_success_samples(samples: list[AlertChainEvidenceSample]) -> None:
+    """Write durable last-success timestamps into the existing Prometheus gauge.
+
+    Do not write `awoooi_alert_chain_healthy` here: that gauge represents an
+    explicit runtime failure signal. Durable success evidence should not mask a
+    later in-memory failure from the same process.
+    """
+    for sample in samples:
+        ALERT_CHAIN_LAST_SUCCESS.labels(source=sample.source).set(
+            sample.last_success_timestamp
+        )
+
+
+def merge_alert_chain_evidence_rows(
+    rows: list[Mapping[str, Any]] | list[Any],
+) -> list[AlertChainEvidenceSample]:
+    """Normalize SQL rows and keep the newest evidence per source."""
+    latest_by_source: dict[str, AlertChainEvidenceSample] = {}
+
+    for row in rows:
+        source = str(_row_value(row, "source") or "").strip().lower()
+        evidence_source = str(_row_value(row, "evidence_source") or "db").strip()
+        raw_timestamp = _row_value(row, "last_success_timestamp")
+        if not source or raw_timestamp is None:
+            continue
+
+        try:
+            timestamp = float(raw_timestamp)
+        except (TypeError, ValueError):
+            continue
+        if timestamp <= 0:
+            continue
+
+        sample = AlertChainEvidenceSample(
+            source=source,
+            last_success_timestamp=timestamp,
+            evidence_source=evidence_source,
+        )
+        previous = latest_by_source.get(source)
+        if previous is None or sample.last_success_timestamp > previous.last_success_timestamp:
+            latest_by_source[source] = sample
+
+    return sorted(latest_by_source.values(), key=lambda item: item.source)
+
+
+def _row_value(row: Mapping[str, Any] | Any, key: str) -> Any:
+    if isinstance(row, Mapping):
+        return row.get(key)
+    return getattr(row, key, None)
+
+
+_ALERT_CHAIN_EVIDENCE_SQL = """
+    WITH conversation_evidence AS (
+        SELECT
+            lower(platform_subject_id) AS source,
+            EXTRACT(EPOCH FROM MAX(COALESCE(provider_ts, received_at))) AS last_success_timestamp,
+            'awooop_conversation_event' AS evidence_source
+        FROM awooop_conversation_event
+        WHERE project_id = :project_id
+          AND channel_type = 'internal'
+          AND lower(COALESCE(platform_subject_id, '')) IN ('alertmanager', 'sentry', 'signoz')
+        GROUP BY lower(platform_subject_id)
+    ),
+    alertmanager_log_evidence AS (
+        SELECT
+            'alertmanager' AS source,
+            EXTRACT(EPOCH FROM MAX(created_at)) AS last_success_timestamp,
+            'alert_operation_log' AS evidence_source
+        FROM alert_operation_log
+        WHERE event_type = 'ALERT_RECEIVED'
+          AND actor = 'alertmanager'
+    )
+    SELECT source, last_success_timestamp, evidence_source
+    FROM conversation_evidence
+    WHERE last_success_timestamp IS NOT NULL
+    UNION ALL
+    SELECT source, last_success_timestamp, evidence_source
+    FROM alertmanager_log_evidence
+    WHERE last_success_timestamp IS NOT NULL
+"""
+
+
+_alert_chain_metrics_service: AlertChainMetricsService | None = None
+
+
+def get_alert_chain_metrics_service() -> AlertChainMetricsService:
+    global _alert_chain_metrics_service
+    if _alert_chain_metrics_service is None:
+        _alert_chain_metrics_service = AlertChainMetricsService()
+    return _alert_chain_metrics_service

apps/api/src/services/alert_rule_engine.py

@@ -33,6 +33,7 @@ import yaml
 
 from src.constants.alert_types import ALERTNAME_TO_TYPE
 from src.services.action_parser import parse_kubectl_action
+from src.services.ollama_endpoint_resolver import resolve_ollama_order
 
 logger = structlog.get_logger(__name__)
 
@@ -297,6 +298,12 @@ def _matches(rule: dict, alertname: str, alert_type: str, message: str, instance
     if alertnames and alertname in alertnames:
         return True
 
+    # 2026-05-31 ogt + Codex: 有明確 alertname 的規則不得只靠寬鬆 message
+    # keyword 命中，否則 HostPreviousBootStorageErrorsDetected 這類主機 storage
+    # 告警會誤配到 minio_disk_high。
+    if alertnames and alertname and alertname != "custom":
+        return False
+
     # alert_type 部分匹配
     for kw in match.get("alert_type", []):
         if kw.lower() in alert_type.lower():
@@ -634,10 +641,12 @@ async def _insert_catalog_ai_generated(
         )
         return
 
-    from sqlalchemy import text as _sql
-    import src.db.base as _db_base
     import json as _json
 
+    from sqlalchemy import text as _sql
+
+    import src.db.base as _db_base
+
     # 從 rule_dict 提取欄位
     # 'expr' 在 OpenClaw YAML 規則中不存在（非 PromQL），
     # 使用 alertname 作為語意佔位（與 yaml_hardcoded 同等策略）
@@ -704,17 +713,37 @@ async def _insert_catalog_ai_generated(
 
 async def _call_ollama(prompt: str, ollama_url: str, model: str) -> str | None:
     """呼叫 Ollama 生成規則 YAML"""
-    try:
-        async with httpx.AsyncClient(timeout=60) as client:
-            resp = await client.post(
-                f"{ollama_url}/api/generate",
-                json={"model": model, "prompt": prompt, "stream": False, "options": {"temperature": 0.1}},
-            )
-            resp.raise_for_status()
-            return resp.json().get("response", "")
-    except Exception as e:
-        logger.warning("auto_rule_ollama_failed", error=str(e))
-        return None
+    endpoints = list(resolve_ollama_order("deep_rca"))
+    if ollama_url and all(endpoint.url != ollama_url for endpoint in endpoints):
+        from types import SimpleNamespace
+
+        endpoints.insert(0, SimpleNamespace(url=ollama_url, provider_name="custom"))
+
+    async with httpx.AsyncClient(timeout=60) as client:
+        for endpoint in endpoints:
+            if not endpoint.url:
+                continue
+            try:
+                resp = await client.post(
+                    f"{endpoint.url}/api/generate",
+                    json={
+                        "model": model,
+                        "prompt": prompt,
+                        "stream": False,
+                        # Preserve response-body compatibility with Ollama 0.24 thinking models.
+                        "think": False,
+                        "options": {"temperature": 0.1},
+                    },
+                )
+                resp.raise_for_status()
+                return resp.json().get("response", "")
+            except Exception as e:
+                logger.warning(
+                    "auto_rule_ollama_failed",
+                    provider=endpoint.provider_name,
+                    error=str(e),
+                )
+    return None
 
 
 async def _call_gemini(prompt: str, api_key: str) -> str | None:

apps/api/src/services/approval_action_classifier.py

@@ -0,0 +1,26 @@
+"""
+Approval action classifier
+==========================
+
+2026-05-31 ogt + Codex: Telegram 告警鏈路一致性修復。
+將 OBSERVE / INVESTIGATE / NO_ACTION 這類「純觀察、未執行修復」的
+判斷集中，避免 execution、Telegram、統計各自用不同語意。
+"""
+
+from __future__ import annotations
+
+
+def is_no_action_approval_action(action: str | None) -> bool:
+    """Return True when an approval action records observation instead of repair."""
+    text = (action or "").strip()
+    upper = text.upper()
+    if not text:
+        return True
+    return (
+        "NO_ACTION" in upper
+        or "NO-ACTION" in upper
+        or "NOACTION" in upper
+        or "(未設)" in text
+        or upper.startswith("OBSERVE")
+        or upper.startswith("INVESTIGATE")
+    )

apps/api/src/services/approval_db.py

@@ -110,10 +110,10 @@ def approval_record_to_request(record: ApprovalRecord) -> ApprovalRequest:
         hit_count=record.hit_count,
         last_seen_at=record.last_seen_at,
         # B3 fix 2026-04-24 ogt + Claude Sonnet 4.6: 補回 DB 欄位（人工審核路徑讀回必要）
-        # incident_id / matched_playbook_id 在 ApprovalRequest 基礎模型中有定義
-        # telegram_message_id / telegram_chat_id 只在 DB model，不在 Pydantic ApprovalRequest
         incident_id=getattr(record, "incident_id", None),
         matched_playbook_id=getattr(record, "matched_playbook_id", None),
+        telegram_message_id=getattr(record, "telegram_message_id", None),
+        telegram_chat_id=getattr(record, "telegram_chat_id", None),
     )
 
 
@@ -659,6 +659,9 @@ class ApprovalDBService:
         approval_id: UUID,
         success: bool,
         error_message: str | None = None,
+        execution_kind: str | None = None,
+        repair_executed: bool | None = None,
+        repair_attempted: bool | None = None,
     ) -> None:
         """
         更新執行狀態
@@ -669,21 +672,47 @@ class ApprovalDBService:
         """
         async with get_db_context() as db:
             status = ApprovalStatus.EXECUTION_SUCCESS if success else ApprovalStatus.EXECUTION_FAILED
-            values: dict = {"status": status}
+            result = await db.execute(
+                select(ApprovalRecord).where(ApprovalRecord.id == str(approval_id))
+            )
+            record = result.scalar_one_or_none()
+            if record is None:
+                logger.warning(
+                    "approval_execution_status_update_missing",
+                    id=str(approval_id),
+                    success=success,
+                )
+                return
+
+            record.status = status
             if not success and error_message:
                 # 截斷至合理長度,避免爆欄位
-                values["rejection_reason"] = str(error_message)[:2000]
-            await db.execute(
-                update(ApprovalRecord)
-                .where(ApprovalRecord.id == str(approval_id))
-                .values(**values)
-            )
+                record.rejection_reason = str(error_message)[:2000]
+            if execution_kind:
+                # 2026-05-31 ogt + Codex: OBSERVE/NO_ACTION 仍需 terminal 狀態，
+                # 但前台/報表必須能分辨「未執行修復」而非真正 execution success。
+                metadata = dict(record.extra_metadata or {})
+                metadata["execution_kind"] = execution_kind
+                metadata["repair_executed"] = (
+                    repair_executed
+                    if repair_executed is not None
+                    else execution_kind not in {
+                        "no_action",
+                        "diagnostic",
+                        "parse_failed",
+                        "unsupported_action",
+                    }
+                )
+                if repair_attempted is not None:
+                    metadata["repair_attempted"] = repair_attempted
+                record.extra_metadata = metadata
 
             logger.info(
                 "approval_execution_status_updated",
                 id=str(approval_id),
                 success=success,
                 has_error=bool(error_message),
+                execution_kind=execution_kind,
             )
 
     async def update_incident_id(self, approval_id: UUID, incident_id: str) -> None:

apps/api/src/services/auto_repair_service.py

@@ -22,9 +22,10 @@ Phase 8: 自動化層實作
 - P0/P1 嚴重度 Incident 需要人工確認
 """
 
-from dataclasses import dataclass
 from collections.abc import Callable
-from typing import Protocol
+from dataclasses import dataclass
+import re
+from typing import Any, Protocol
 
 import structlog
 
@@ -81,6 +82,55 @@ class AutoRepairResult:
     execution_time_ms: int = 0
 
 
+@dataclass(frozen=True)
+class _SshMcpRoute:
+    """Route a legacy SSH playbook command to a governed MCP tool."""
+
+    tool_name: str
+    params: dict[str, Any]
+
+
+_SHORT_HOST_MAP: dict[str, str] = {
+    "110": "192.168.0.110",
+    "120": "192.168.0.120",
+    "121": "192.168.0.121",
+    "188": "192.168.0.188",
+}
+
+_SSH_DIAGNOSTIC_KEYWORDS = (
+    "ps aux",
+    "docker stats",
+    "docker inspect",
+    "docker logs",
+    "docker ps",
+    "docker top",
+    "df -h",
+    "du -",
+    "free -h",
+    "journalctl",
+    "systemctl show",
+    "tail ",
+    "top ",
+    "uptime",
+)
+
+_SSH_WRITE_KEYWORDS = (
+    "docker restart",
+    "docker start",
+    "docker stop",
+    "docker rm",
+    "docker prune",
+    "systemctl restart",
+    "systemctl stop",
+    "systemctl start",
+    "truncate ",
+    " rm ",
+    "rm -",
+    "certbot renew",
+    "bash ",
+)
+
+
 # =============================================================================
 # Auto Repair Service Interface
 # =============================================================================
@@ -108,6 +158,7 @@ class IAutoRepairService(Protocol):
         self,
         incident: Incident,
         playbook: Playbook,
+        run_post_verification: bool = True,
     ) -> AutoRepairResult:
         """
         執行自動修復
@@ -320,7 +371,16 @@ class AutoRepairService:
             )
 
         # 4. 檢查最佳匹配
-        best_match = recommendations[0]
+        best_match = self._select_best_recommendation(recommendations, symptoms)
+        if best_match is not recommendations[0]:
+            logger.warning(
+                "auto_repair_exact_match_prioritized",
+                incident_id=incident.incident_id,
+                selected_playbook_id=best_match.playbook.playbook_id,
+                original_playbook_id=recommendations[0].playbook.playbook_id,
+                selected_similarity=best_match.similarity_score,
+                original_similarity=recommendations[0].similarity_score,
+            )
 
         # 2026-04-07 Claude Code: 統帥指令「直接全部跳成自動修復」
         # 移除: 相似度門檻、is_high_quality 門檻、冷啟動機制、風險等級門檻
@@ -378,6 +438,7 @@ class AutoRepairService:
         playbook: Playbook,
         is_cold_start: bool = False,
         similarity_score: float | None = None,
+        run_post_verification: bool = True,
     ) -> AutoRepairResult:
         """
         執行自動修復
@@ -414,6 +475,8 @@ class AutoRepairService:
                 executed_steps.append(
                     f"Step {step.step_number}: {step.command[:50]}... -> {step_result}"
                 )
+                if self._is_step_failure_result(step_result):
+                    raise RuntimeError(f"Step {step.step_number} failed: {step_result}")
 
             # 更新 Playbook 統計
             await self._playbook_service.record_execution(
@@ -579,10 +642,17 @@ class AutoRepairService:
                             error=str(_inner_e),
                         )
 
-                _vl_task = _asyncio.create_task(_verify_and_learn())
-                if hasattr(self, "_pending_tasks"):
-                    self._pending_tasks.add(_vl_task)
-                    _vl_task.add_done_callback(self._pending_tasks.discard)
+                if run_post_verification:
+                    _vl_task = _asyncio.create_task(_verify_and_learn())
+                    if hasattr(self, "_pending_tasks"):
+                        self._pending_tasks.add(_vl_task)
+                        _vl_task.add_done_callback(self._pending_tasks.discard)
+                else:
+                    logger.info(
+                        "auto_repair_service_post_verify_delegated",
+                        incident_id=incident.incident_id,
+                        playbook_id=playbook.playbook_id,
+                    )
             except Exception as _vl_e:
                 logger.warning("auto_repair_verifier_setup_failed", error=str(_vl_e))
 
@@ -688,6 +758,44 @@ class AutoRepairService:
             keywords=keywords[:10],
         )
 
+    def _select_best_recommendation(
+        self,
+        recommendations,
+        symptoms: SymptomPattern,
+    ):
+        """Prefer deterministic alert/service matches over fuzzy similarity only.
+
+        A higher fuzzy score must not outrank a playbook that explicitly names the
+        firing alert or affected service. Live-fire T16 proved that this can route
+        a safe K8s canary into an unrelated host diagnostic playbook.
+        """
+
+        symptom_alerts = {str(name) for name in (symptoms.alert_names or []) if name}
+        symptom_services = {
+            str(service) for service in (symptoms.affected_services or []) if service
+        }
+
+        def _priority(recommendation) -> tuple[int, int, float]:
+            pattern = recommendation.playbook.symptom_pattern
+            playbook_alerts = {
+                str(name) for name in (pattern.alert_names or []) if name
+            }
+            playbook_services = {
+                str(service) for service in (pattern.affected_services or []) if service
+            }
+            alert_exact = int(bool(symptom_alerts & playbook_alerts))
+            service_exact = int(bool(symptom_services & playbook_services))
+            return (alert_exact, service_exact, float(recommendation.similarity_score or 0.0))
+
+        return max(recommendations, key=_priority)
+
+    @staticmethod
+    def _is_step_failure_result(step_result: str) -> bool:
+        """Treat executor-declared failures as failed auto-repair executions."""
+
+        normalized = (step_result or "").strip().upper()
+        return normalized.startswith("FAILED:") or normalized == "UNKNOWN_ACTION_TYPE"
+
     def _get_max_risk_level(self, playbook: Playbook) -> RiskLevel:
         """取得 Playbook 中最高的風險等級"""
         risk_order = {
@@ -860,6 +968,175 @@ class AutoRepairService:
             # 安全降級：檢查失敗 → 保守拒絕
             return False
 
+    def _route_legacy_ssh_command_to_mcp(
+        self,
+        incident: Incident,
+        command: str,
+    ) -> _SshMcpRoute | None:
+        """Map read-only legacy ``ssh {host} '...'`` steps to MCP Gateway.
+
+        YAML_RULE playbooks predate the URI executor and can contain compound
+        shell diagnostics.  Those commands should not bypass the newer
+        scheme-based HostRepairAgent or loosen its shell safety guard; read-only
+        diagnostics are instead routed to the governed SSH MCP provider.
+        """
+
+        raw_command = (command or "").strip()
+        lowered = raw_command.lower()
+        if not lowered.startswith("ssh "):
+            return None
+
+        if any(token in lowered for token in _SSH_WRITE_KEYWORDS):
+            return None
+
+        if not any(token in lowered for token in _SSH_DIAGNOSTIC_KEYWORDS):
+            return None
+
+        host = self._resolve_ssh_host_for_incident(incident, raw_command)
+        if not host:
+            return None
+
+        params: dict[str, Any] = {"host": host}
+        container_name = self._resolve_container_name_for_incident(incident, raw_command)
+        if container_name:
+            params["container_name"] = container_name
+
+        return _SshMcpRoute(tool_name="ssh_diagnose", params=params)
+
+    def preview_read_only_ssh_mcp_route(
+        self,
+        incident: Incident,
+        command: str,
+    ) -> dict[str, Any] | None:
+        """Preview whether a legacy SSH diagnostic can use the MCP Gateway.
+
+        This is used by remediation dry-runs to prove the supported executor
+        path without running the original PlayBook step or writing an execution
+        result.
+        """
+
+        route = self._route_legacy_ssh_command_to_mcp(incident, command)
+        if route is None:
+            return None
+        return {
+            "tool_name": route.tool_name,
+            "params": route.params,
+            "agent_id": "auto_repair_executor",
+            "required_scope": "read",
+            "flywheel_node": "execute",
+        }
+
+    def _resolve_ssh_host_for_incident(self, incident: Incident, command: str) -> str:
+        """Resolve ``{host}``, short host labels, and exporter instance ports."""
+
+        labels = self._incident_labels(incident)
+        raw_host = ""
+        match = re.match(r"ssh\s+([^\s'\"]+)", command.strip(), flags=re.IGNORECASE)
+        if match:
+            raw_host = match.group(1)
+
+        if not raw_host or "{" in raw_host or "}" in raw_host:
+            raw_host = (
+                str(labels.get("host") or "")
+                or str(labels.get("instance") or "")
+                or str(labels.get("node") or "")
+                or str(labels.get("exported_instance") or "")
+            )
+
+        return self._normalize_ssh_host(raw_host)
+
+    @staticmethod
+    def _normalize_ssh_host(raw_host: str) -> str:
+        host = (raw_host or "").strip()
+        if host.startswith("ssh://"):
+            host = host.removeprefix("ssh://")
+        if "@" in host:
+            host = host.rsplit("@", 1)[1]
+        if host.startswith("[") and "]" in host:
+            host = host[1:host.index("]")]
+        if host.count(":") == 1:
+            maybe_host, maybe_port = host.rsplit(":", 1)
+            if maybe_port.isdigit():
+                host = maybe_host
+        if host in _SHORT_HOST_MAP:
+            return _SHORT_HOST_MAP[host]
+        match = re.fullmatch(r"(?:node-exporter-|host-)?(110|120|121|188)", host)
+        if match:
+            return _SHORT_HOST_MAP[match.group(1)]
+        return host
+
+    def _resolve_container_name_for_incident(
+        self,
+        incident: Incident,
+        command: str,
+    ) -> str:
+        labels = self._incident_labels(incident)
+        for key in ("container_name", "container", "name"):
+            value = str(labels.get(key) or "").strip()
+            if value and "{" not in value and "}" not in value:
+                return value
+
+        match = re.search(
+            r"docker\s+(?:stats\s+--no-stream|inspect|logs|top|ps\s+-a\s+--filter\s+name=)\s+([a-zA-Z0-9._-]+)",
+            command,
+        )
+        return match.group(1) if match else ""
+
+    @staticmethod
+    def _incident_labels(incident: Incident) -> dict[str, Any]:
+        for signal in incident.signals or []:
+            labels = getattr(signal, "labels", None)
+            if labels:
+                return labels
+        return {}
+
+    async def _execute_ssh_mcp_route(
+        self,
+        incident: Incident,
+        route: _SshMcpRoute,
+    ) -> str:
+        """Execute a routed SSH diagnostic through AwoooP MCP Gateway."""
+
+        try:
+            from src.db.base import get_db_context
+            from src.plugins.mcp.gateway import GatewayContext, McpGateway, McpGatewayError
+            from src.services.mcp_audit_context import with_mcp_audit_context
+
+            incident_id = incident.incident_id
+            params = with_mcp_audit_context(
+                route.params,
+                session_id=f"incident:{incident_id}:auto_repair_execute",
+                incident_id=incident_id,
+                flywheel_node="execute",
+                agent_role="auto_repair_executor",
+            )
+            async with get_db_context("awoooi") as db:
+                ctx = GatewayContext(
+                    project_id="awoooi",
+                    agent_id="auto_repair_executor",
+                    tool_name=route.tool_name,
+                    trace_id=incident_id,
+                    is_shadow=False,
+                    environment={"env": "prod"},
+                    required_scope="read",
+                )
+                result = await McpGateway(db).call(ctx, params)
+        except McpGatewayError as exc:
+            return f"FAILED: mcp:{route.tool_name} {exc.error_code}: {exc}"
+        except Exception as exc:
+            logger.warning(
+                "auto_repair_ssh_mcp_route_failed",
+                incident_id=incident.incident_id,
+                tool=route.tool_name,
+                error=str(exc),
+            )
+            return f"FAILED: mcp:{route.tool_name} {exc}"
+
+        if result.success:
+            preview = str(result.output or "")[:500]
+            return f"SUCCESS: mcp:{route.tool_name} {preview}".strip()
+        return f"FAILED: mcp:{route.tool_name} {result.error or 'execution failed'}"
+
     async def _execute_step(self, incident: Incident, step) -> str:
         """
         執行單一修復步驟
@@ -891,6 +1168,10 @@ class AutoRepairService:
 
         # 2026-04-06 Claude Code: Sprint 3 — repair_by_uri (URI scheme 路由)
         if step.action_type == ActionType.SSH_COMMAND:
+            route = self._route_legacy_ssh_command_to_mcp(incident, step.command)
+            if route is not None:
+                return await self._execute_ssh_mcp_route(incident, route)
+
             from src.services.host_repair_agent import HostRepairAgent
             agent = HostRepairAgent()
             approved = not getattr(step, "requires_approval", False)

apps/api/src/services/awooop_ansible_audit_service.py

@@ -0,0 +1,633 @@
+"""AwoooP Ansible audit helpers.
+
+This module is intentionally non-executing. It exposes the Ansible audit
+contract and repo-known playbook catalog so the truth chain can say whether
+Ansible was actually considered or executed, without pretending that catalog
+hints are runtime remediation.
+"""
+
+from __future__ import annotations
+
+import json
+from typing import Any
+
+import structlog
+from sqlalchemy import text
+
+from src.db.base import get_db_context
+
+logger = structlog.get_logger(__name__)
+
+
+def _automation_operation_log_incident_id(value: str) -> int | None:
+    # ADR-090 keeps external INC-* ids in input JSON; the DB column is BIGINT.
+    normalized = str(value or "").strip()
+    return int(normalized) if normalized.isdigit() else None
+
+
+ANSIBLE_OPERATION_TYPES = frozenset({
+    "ansible_candidate_matched",
+    "ansible_check_mode_executed",
+    "ansible_apply_executed",
+    "ansible_rollback_executed",
+    "ansible_execution_skipped",
+})
+
+_CATALOG: tuple[dict[str, Any], ...] = (
+    {
+        "catalog_id": "ansible:110-devops",
+        "playbook_path": "infra/ansible/playbooks/110-devops.yml",
+        "inventory_hosts": ["host_110"],
+        "domains": ["swap", "harbor", "sentry", "gitea", "langfuse", "bitan", "runner", "keepalived", "nginx"],
+        "keywords": [
+            "110",
+            "docker",
+            "container",
+            "dockercontainerunhealthy",
+            "swap",
+            "harbor",
+            "sentry",
+            "gitea",
+            "langfuse",
+            "bitan",
+            "runner",
+            "github-runner",
+            "keepalived",
+        ],
+        "supports_check_mode": True,
+        "auto_apply_enabled": False,
+        "approval_required": True,
+        "risk_level": "medium",
+    },
+    {
+        "catalog_id": "ansible:188-momo-backup-user",
+        "playbook_path": "infra/ansible/playbooks/188-momo-backup-user.yml",
+        "check_mode_playbook_path": "infra/ansible/playbooks/188-momo-backup-user.yml",
+        "inventory_hosts": ["host_188"],
+        "domains": ["momo_backup", "postgresql", "cron", "awooop_notification"],
+        "keywords": [
+            "188",
+            "momo",
+            "momopostgresbackupfailed",
+            "postgres",
+            "postgresql",
+            "backup",
+            "momo pg_backup",
+            "momo postgres backup",
+            "pg_backup",
+            "momo-pg-backup",
+            "cron",
+            "crontab",
+        ],
+        "supports_check_mode": True,
+        "auto_apply_enabled": False,
+        "approval_required": True,
+        "risk_level": "low",
+    },
+    {
+        "catalog_id": "ansible:188-ai-web",
+        "playbook_path": "infra/ansible/playbooks/188-ai-web.yml",
+        "check_mode_playbook_path": "infra/ansible/playbooks/188-ai-web-readonly.yml",
+        "inventory_hosts": ["host_188"],
+        "domains": ["docker", "momo_backup", "signoz", "minio", "litellm", "n8n", "open_webui", "nginx"],
+        "keywords": [
+            "188",
+            "docker",
+            "container",
+            "dockercontainerunhealthy",
+            "momo",
+            "backup",
+            "postgresql",
+            "pg_backup",
+            "signoz",
+            "minio",
+            "litellm",
+            "n8n",
+            "open-webui",
+            "openwebui",
+            "docker-registry",
+        ],
+        "supports_check_mode": True,
+        "auto_apply_enabled": False,
+        "approval_required": True,
+        "risk_level": "medium",
+    },
+    {
+        "catalog_id": "ansible:nginx-sync",
+        "playbook_path": "infra/ansible/playbooks/nginx-sync.yml",
+        "inventory_hosts": ["host_110", "host_188"],
+        "domains": ["nginx", "proxy", "ollama_proxy", "tls"],
+        "keywords": ["nginx", "proxy", "ollama", "gcp", "tls", "cert", "502", "upstream"],
+        "supports_check_mode": True,
+        "auto_apply_enabled": False,
+        "approval_required": True,
+        "risk_level": "medium",
+    },
+    {
+        "catalog_id": "ansible:restore-password-auth",
+        "playbook_path": "infra/ansible/playbooks/restore-password-auth.yml",
+        "inventory_hosts": ["host_110", "host_120", "host_121", "host_188"],
+        "domains": ["ssh", "password_auth"],
+        "keywords": ["ssh", "passwordauthentication", "password auth", "login", "auth"],
+        "supports_check_mode": False,
+        "auto_apply_enabled": False,
+        "approval_required": True,
+        "risk_level": "high",
+    },
+)
+
+
+def get_ansible_catalog_item(catalog_id: str) -> dict[str, Any] | None:
+    """Return one repo-known Ansible catalog item without exposing mutability."""
+
+    for item in _CATALOG:
+        if item["catalog_id"] == catalog_id:
+            return dict(item)
+    return None
+
+
+def _get(row: dict[str, Any], key: str) -> Any:
+    return row.get(key)
+
+
+def _tags(row: dict[str, Any]) -> list[str]:
+    raw = _get(row, "tags")
+    if isinstance(raw, list):
+        return [str(item).lower() for item in raw]
+    if isinstance(raw, str):
+        return [part.strip().lower() for part in raw.split(",") if part.strip()]
+    return []
+
+
+def _first_present(row: dict[str, Any], keys: tuple[str, ...]) -> Any:
+    for key in keys:
+        value = _get(row, key)
+        if value not in (None, ""):
+            return value
+    return None
+
+
+def _json_object(row: dict[str, Any], key: str) -> dict[str, Any]:
+    raw = _get(row, key)
+    if isinstance(raw, dict):
+        return raw
+    if isinstance(raw, str) and raw.strip():
+        try:
+            parsed = json.loads(raw)
+        except json.JSONDecodeError:
+            return {}
+        return parsed if isinstance(parsed, dict) else {}
+    return {}
+
+
+def _json_value(row: dict[str, Any], container_key: str, value_key: str) -> Any:
+    return _json_object(row, container_key).get(value_key)
+
+
+def _first_present_with_json(row: dict[str, Any], keys: tuple[str, ...], json_keys: tuple[tuple[str, str], ...] = ()) -> Any:
+    value = _first_present(row, keys)
+    if value not in (None, ""):
+        return value
+    for container_key, value_key in json_keys:
+        value = _json_value(row, container_key, value_key)
+        if value not in (None, ""):
+            return value
+    return None
+
+
+def _bool_or_none(value: Any) -> bool | None:
+    if isinstance(value, bool):
+        return value
+    if value in (None, ""):
+        return None
+    normalized = str(value).strip().lower()
+    if normalized in {"1", "true", "yes", "y", "on"}:
+        return True
+    if normalized in {"0", "false", "no", "n", "off"}:
+        return False
+    return None
+
+
+def _int_or_none(value: Any) -> int | None:
+    if value in (None, ""):
+        return None
+    try:
+        return int(value)
+    except (TypeError, ValueError):
+        return None
+
+
+def _is_ansible_operation(row: dict[str, Any]) -> bool:
+    operation_type = str(_get(row, "operation_type") or "").lower()
+    if operation_type in ANSIBLE_OPERATION_TYPES:
+        return True
+    if "ansible" in _tags(row):
+        return True
+    executor = str(
+        _first_present(
+            row,
+            (
+                "input_executor",
+                "input_execution_backend",
+                "output_executor",
+                "output_execution_backend",
+            ),
+        )
+        or ""
+    ).lower()
+    if executor == "ansible":
+        return True
+    playbook_path = str(
+        _first_present(row, ("input_playbook_path", "output_playbook_path", "input_ansible_playbook_path", "output_ansible_playbook_path"))
+        or ""
+    ).lower()
+    return "infra/ansible/" in playbook_path or playbook_path.endswith(".yml") and "ansible" in playbook_path
+
+
+def _ansible_record(row: dict[str, Any]) -> dict[str, Any]:
+    return {
+        "op_id": _get(row, "op_id"),
+        "parent_op_id": _get(row, "parent_op_id"),
+        "operation_type": _get(row, "operation_type"),
+        "status": _get(row, "status"),
+        "actor": _get(row, "actor"),
+        "catalog_id": _first_present(row, ("input_catalog_id", "output_catalog_id")),
+        "playbook_id": _first_present(row, ("input_playbook_id", "output_playbook_id")),
+        "playbook_path": _first_present(
+            row,
+            ("input_playbook_path", "output_playbook_path", "input_ansible_playbook_path", "output_ansible_playbook_path"),
+        ),
+        "execution_mode": _first_present(row, ("input_execution_mode", "output_execution_mode")),
+        "check_mode": _first_present(row, ("input_check_mode", "output_check_mode")),
+        "check_mode_executed": _first_present_with_json(
+            row,
+            ("input_check_mode_executed", "output_check_mode_executed", "dry_run_check_mode_executed"),
+            (("dry_run_result", "check_mode_executed"),),
+        ),
+        "apply_enabled": _first_present(row, ("input_apply_enabled", "output_apply_enabled")),
+        "apply_executed": _first_present_with_json(
+            row,
+            ("input_apply_executed", "output_apply_executed", "dry_run_apply_executed"),
+            (("dry_run_result", "apply_executed"),),
+        ),
+        "approval_source": _first_present(row, ("input_approval_source", "output_approval_source")),
+        "returncode": _first_present_with_json(
+            row,
+            ("output_returncode", "input_returncode", "dry_run_returncode"),
+            (("dry_run_result", "returncode"),),
+        ),
+        "not_used_reason": _first_present(row, ("input_not_used_reason", "output_not_used_reason")),
+        "dry_run_result": _get(row, "dry_run_result"),
+        "error": _get(row, "error"),
+        "duration_ms": _get(row, "duration_ms"),
+        "tags": _get(row, "tags"),
+        "created_at": _get(row, "created_at"),
+    }
+
+
+def summarize_ansible_execution(records: list[dict[str, Any]]) -> dict[str, Any]:
+    """Summarize durable Ansible audit rows for Telegram and Operator UI."""
+    typed_records = [row for row in records if isinstance(row, dict)]
+    check_mode_total = 0
+    apply_total = 0
+    rollback_total = 0
+    pending_check_mode_total = 0
+    applied_success_total = 0
+
+    terminal_check_mode_parent_ids = {
+        str(row.get("parent_op_id"))
+        for row in typed_records
+        if str(row.get("operation_type") or "") in {
+            "ansible_check_mode_executed",
+            "ansible_execution_skipped",
+        }
+        and row.get("parent_op_id")
+    }
+
+    for row in typed_records:
+        operation_type = str(row.get("operation_type") or "")
+        status = str(row.get("status") or "").lower()
+        if operation_type == "ansible_check_mode_executed" and status != "pending":
+            check_mode_total += 1
+        elif operation_type == "ansible_apply_executed":
+            apply_total += 1
+            if status in {"success", "completed", "executed", "applied"}:
+                applied_success_total += 1
+        elif operation_type == "ansible_rollback_executed":
+            rollback_total += 1
+        elif (
+            operation_type == "ansible_candidate_matched"
+            and str(row.get("op_id")) not in terminal_check_mode_parent_ids
+        ):
+            pending_check_mode_total += 1
+
+    latest_record = typed_records[0] if typed_records else {}
+    latest_apply = next(
+        (row for row in typed_records if row.get("operation_type") == "ansible_apply_executed"),
+        {},
+    )
+    latest_check = next(
+        (row for row in typed_records if row.get("operation_type") == "ansible_check_mode_executed"),
+        {},
+    )
+    focused = latest_apply or latest_check or latest_record
+    returncode = _int_or_none(focused.get("returncode"))
+    approval_source = focused.get("approval_source")
+
+    return {
+        "check_mode_total": check_mode_total,
+        "apply_total": apply_total,
+        "rollback_total": rollback_total,
+        "pending_check_mode_total": pending_check_mode_total,
+        "applied_success_total": applied_success_total,
+        "applied": applied_success_total > 0,
+        "controlled_apply": bool(latest_apply) and bool(approval_source),
+        "latest_operation_type": focused.get("operation_type"),
+        "latest_status": focused.get("status"),
+        "latest_catalog_id": focused.get("catalog_id"),
+        "latest_playbook_path": focused.get("playbook_path"),
+        "latest_execution_mode": focused.get("execution_mode"),
+        "latest_check_mode": _bool_or_none(focused.get("check_mode")),
+        "latest_check_mode_executed": _bool_or_none(focused.get("check_mode_executed")),
+        "latest_apply_enabled": _bool_or_none(focused.get("apply_enabled")),
+        "latest_apply_executed": _bool_or_none(focused.get("apply_executed")),
+        "latest_returncode": returncode if returncode is not None else focused.get("returncode"),
+        "approval_source": approval_source,
+        "latest_actor": focused.get("actor"),
+        "latest_op_id": focused.get("op_id"),
+        "latest_parent_op_id": focused.get("parent_op_id"),
+    }
+
+
+def _flatten_text(value: Any, pieces: list[str], remaining: int = 80) -> int:
+    if remaining <= 0 or value is None:
+        return remaining
+    if isinstance(value, dict):
+        for key, item in value.items():
+            remaining = _flatten_text(key, pieces, remaining)
+            remaining = _flatten_text(item, pieces, remaining)
+            if remaining <= 0:
+                break
+        return remaining
+    if isinstance(value, list):
+        for item in value:
+            remaining = _flatten_text(item, pieces, remaining)
+            if remaining <= 0:
+                break
+        return remaining
+    pieces.append(str(value).lower())
+    return remaining - 1
+
+
+def _source_haystack(incident: dict[str, Any] | None, drift: dict[str, Any] | None) -> str:
+    pieces: list[str] = []
+    _flatten_text(incident, pieces)
+    _flatten_text(drift, pieces)
+    return " ".join(pieces)
+
+
+def _catalog_hints(incident: dict[str, Any] | None, drift: dict[str, Any] | None) -> dict[str, Any]:
+    haystack = _source_haystack(incident, drift)
+    candidates: list[dict[str, Any]] = []
+    unmatched: list[str] = []
+    for item in _CATALOG:
+        matched = [keyword for keyword in item["keywords"] if keyword in haystack]
+        public_item = {
+            key: value
+            for key, value in item.items()
+            if key
+            in {
+                "catalog_id",
+                "playbook_path",
+                "inventory_hosts",
+                "domains",
+                "supports_check_mode",
+                "check_mode_playbook_path",
+                "auto_apply_enabled",
+                "approval_required",
+                "risk_level",
+            }
+        }
+        if matched:
+            candidates.append({
+                **public_item,
+                "match_score": len(matched),
+                "matched_keywords": matched,
+            })
+        else:
+            unmatched.append(item["catalog_id"])
+    candidates.sort(key=lambda row: (-int(row["match_score"]), str(row["catalog_id"])))
+    return {
+        "match_mode": "static_catalog_keyword_hint_v1",
+        "decision_effect": "none",
+        "available_count": len(_CATALOG),
+        "candidates": candidates,
+        "unmatched_catalog_ids": unmatched,
+    }
+
+
+def build_ansible_truth(
+    automation_ops: list[dict[str, Any]],
+    *,
+    incident: dict[str, Any] | None,
+    drift: dict[str, Any] | None,
+) -> dict[str, Any]:
+    """Build the truth-chain Ansible section from audited facts and catalog hints."""
+
+    records = [_ansible_record(row) for row in automation_ops if _is_ansible_operation(row)]
+    summary = summarize_ansible_execution(records)
+    return {
+        "considered": bool(records),
+        "records": records,
+        "summary": summary,
+        "audit_contract": {
+            "schema_version": "ansible_executor_audit_v1",
+            "operation_types": sorted(ANSIBLE_OPERATION_TYPES),
+            "required_audit_fields": [
+                "incident_id",
+                "operation_type",
+                "status",
+                "actor",
+                "input.executor",
+                "input.catalog_id",
+                "input.execution_mode",
+                "input.playbook_path",
+                "input.check_mode",
+                "input.approval_source",
+                "output.returncode",
+                "output.not_used_reason",
+                "dry_run_result",
+            ],
+            "default_execution_mode": "catalog/dry-run audit only until approval execution is explicitly wired",
+        },
+        "candidate_catalog": _catalog_hints(incident, drift),
+        "not_used_reason": (
+            None
+            if records
+            else "no automation_operation_log row with Ansible operation type, tag, or executor backend for this source"
+        ),
+    }
+
+
+def _incident_public_dict(incident: Any) -> dict[str, Any]:
+    if incident is None:
+        return {}
+    if isinstance(incident, dict):
+        return incident
+    severity = getattr(incident, "severity", None)
+    signals_payload: list[dict[str, Any]] = []
+    for signal in getattr(incident, "signals", None) or []:
+        signals_payload.append({
+            "alert_name": getattr(signal, "alert_name", None),
+            "labels": getattr(signal, "labels", None) or {},
+            "annotations": getattr(signal, "annotations", None) or {},
+        })
+    return {
+        "incident_id": getattr(incident, "incident_id", None),
+        "project_id": getattr(incident, "project_id", None),
+        "alertname": getattr(incident, "alertname", None),
+        "alert_category": getattr(incident, "alert_category", None),
+        "notification_type": getattr(incident, "notification_type", None),
+        "severity": getattr(severity, "value", severity),
+        "affected_services": getattr(incident, "affected_services", None) or [],
+        "signals": signals_payload,
+    }
+
+
+def build_ansible_decision_audit_payload(
+    *,
+    incident: Any,
+    proposal_data: dict[str, Any],
+    decision_path: str,
+    not_used_reason: str,
+) -> dict[str, Any] | None:
+    """Return an AOL payload when Ansible has catalog candidates for a decision."""
+
+    incident_payload = _incident_public_dict(incident)
+    hints = _catalog_hints(incident_payload, None)
+    candidates = hints.get("candidates") or []
+    if not candidates:
+        return None
+
+    incident_id = str(incident_payload.get("incident_id") or "")
+    input_payload = {
+        "incident_id": incident_id,
+        "executor": "ansible",
+        "execution_backend": "ansible",
+        "decision_path": decision_path,
+        "check_mode": True,
+        "apply_enabled": False,
+        "approval_required": True,
+        "candidate_catalog_schema": hints["match_mode"],
+        "executor_candidates": [
+            {
+                "catalog_id": row["catalog_id"],
+                "playbook_path": row["playbook_path"],
+                "check_mode_playbook_path": row.get("check_mode_playbook_path"),
+                "inventory_hosts": row["inventory_hosts"],
+                "supports_check_mode": row["supports_check_mode"],
+                "auto_apply_enabled": row["auto_apply_enabled"],
+                "approval_required": row["approval_required"],
+                "risk_level": row["risk_level"],
+                "match_score": row["match_score"],
+                "matched_keywords": row["matched_keywords"],
+            }
+            for row in candidates[:5]
+        ],
+        "proposal_source": proposal_data.get("source", ""),
+        "proposal_risk_level": proposal_data.get("risk_level", ""),
+        "proposal_action_preview": str(
+            proposal_data.get("action")
+            or proposal_data.get("kubectl_command")
+            or ""
+        )[:240],
+    }
+    output_payload = {
+        "not_used_reason": not_used_reason,
+        "decision_effect": "audit_only",
+        "next_required_step": "wire approval_execution to Ansible check-mode before apply",
+    }
+    return {
+        "operation_type": "ansible_candidate_matched",
+        "status": "dry_run",
+        "input": input_payload,
+        "output": output_payload,
+        "dry_run_result": {
+            "check_mode_executed": False,
+            "candidate_count": len(candidates),
+            "reason": not_used_reason,
+        },
+        "tags": ["ansible", "decision", "candidate", "check_mode_pending"],
+    }
+
+
+async def record_ansible_decision_audit(
+    *,
+    incident: Any,
+    proposal_data: dict[str, Any],
+    decision_path: str,
+    not_used_reason: str,
+) -> bool:
+    """Write a best-effort Ansible candidate audit row for one decision."""
+
+    payload = build_ansible_decision_audit_payload(
+        incident=incident,
+        proposal_data=proposal_data,
+        decision_path=decision_path,
+        not_used_reason=not_used_reason,
+    )
+    if payload is None:
+        return False
+
+    incident_id = payload["input"]["incident_id"]
+    project_id = getattr(incident, "project_id", None) or "awoooi"
+    try:
+        async with get_db_context(str(project_id)) as db:
+            existing = await db.execute(
+                text("""
+                    SELECT op_id
+                    FROM automation_operation_log
+                    WHERE operation_type = 'ansible_candidate_matched'
+                      AND coalesce(incident_id::text, input ->> 'incident_id') = :incident_id
+                      AND input ->> 'executor' = 'ansible'
+                    LIMIT 1
+                """),
+                {"incident_id": incident_id},
+            )
+            if existing.scalar() is not None:
+                return False
+            await db.execute(
+                text("""
+                    INSERT INTO automation_operation_log (
+                        operation_type, actor, status, incident_id,
+                        input, output, dry_run_result, tags
+                    ) VALUES (
+                        :operation_type,
+                        'decision_manager',
+                        :status,
+                        :incident_db_id,
+                        CAST(:input AS jsonb),
+                        CAST(:output AS jsonb),
+                        CAST(:dry_run_result AS jsonb),
+                        :tags
+                    )
+                """),
+                {
+                    "operation_type": payload["operation_type"],
+                    "status": payload["status"],
+                    "incident_id": incident_id,
+                    "incident_db_id": _automation_operation_log_incident_id(incident_id),
+                    "input": json.dumps(payload["input"], ensure_ascii=False),
+                    "output": json.dumps(payload["output"], ensure_ascii=False),
+                    "dry_run_result": json.dumps(payload["dry_run_result"], ensure_ascii=False),
+                    "tags": payload["tags"],
+                },
+            )
+            return True
+    except Exception as exc:
+        logger.warning(
+            "ansible_decision_audit_write_failed",
+            incident_id=incident_id,
+            error=str(exc),
+        )
+        return False

apps/api/src/services/awooop_ansible_check_mode_service.py

@@ -0,0 +1,661 @@
+"""Safe Ansible check-mode executor for AwoooP truth-chain evidence.
+
+This service is deliberately dry-run only. It claims pending
+``ansible_candidate_matched`` AOL rows, runs ``ansible-playbook --check --diff``,
+and writes the result back as ``ansible_check_mode_executed``. It never enables
+apply and never writes auto_repair_executions.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import os
+import re
+import shutil
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+import structlog
+from sqlalchemy import text
+
+from src.core.config import settings
+from src.db.base import get_db_context
+from src.services.awooop_ansible_audit_service import get_ansible_catalog_item
+
+logger = structlog.get_logger(__name__)
+
+_SAFE_HOST_RE = re.compile(r"^[A-Za-z0-9_.-]+$")
+_PLAYBOOK_PREFIX = Path("infra/ansible/playbooks")
+_STDOUT_LIMIT = 20_000
+_STDERR_LIMIT = 12_000
+FORCED_COMMAND_BLOCKER = "ansible_repair_ssh_forced_command_denies_ansible_bootstrap"
+REPAIR_FORCED_COMMAND_KEY_PATH = Path("/etc/repair-ssh/id_ed25519")
+REPAIR_FORCED_COMMAND_KNOWN_HOSTS_PATH = Path("/etc/repair-known-hosts/known_hosts")
+
+
+@dataclass(frozen=True)
+class AnsibleCheckModeClaim:
+    op_id: str
+    source_candidate_op_id: str
+    incident_id: str
+    catalog_id: str
+    playbook_path: str
+    inventory_hosts: tuple[str, ...]
+    input_payload: dict[str, Any]
+
+
+@dataclass(frozen=True)
+class AnsibleCommandSpec:
+    command: list[str]
+    cwd: Path
+    env: dict[str, str]
+    playbook_abs_path: Path
+    inventory_abs_path: Path
+
+
+@dataclass(frozen=True)
+class AnsibleRunResult:
+    returncode: int
+    stdout: str
+    stderr: str
+    duration_ms: int
+    timed_out: bool = False
+
+
+def _tail(text_value: str, limit: int) -> str:
+    if len(text_value) <= limit:
+        return text_value
+    return text_value[-limit:]
+
+
+def _json_loads(value: Any) -> dict[str, Any]:
+    if isinstance(value, dict):
+        return value
+    if isinstance(value, str):
+        try:
+            parsed = json.loads(value)
+        except json.JSONDecodeError:
+            return {}
+        return parsed if isinstance(parsed, dict) else {}
+    return {}
+
+
+def _incident_id_from_payload(payload: dict[str, Any]) -> str:
+    return str(payload.get("incident_id") or "").strip()
+
+
+def _automation_operation_log_incident_id(value: str) -> int | None:
+    # ADR-090 keeps external INC-* ids in input JSON; the DB column is BIGINT.
+    normalized = str(value or "").strip()
+    return int(normalized) if normalized.isdigit() else None
+
+
+def _check_mode_ssh_key_path() -> Path:
+    return Path(settings.AWOOOP_ANSIBLE_CHECK_MODE_SSH_KEY_PATH)
+
+
+def _check_mode_known_hosts_path() -> Path:
+    return Path(settings.AWOOOP_ANSIBLE_CHECK_MODE_KNOWN_HOSTS_PATH)
+
+
+def _uses_repair_forced_command_transport(key_path: Path | None = None) -> bool:
+    return (key_path or _check_mode_ssh_key_path()) == REPAIR_FORCED_COMMAND_KEY_PATH
+
+
+def detect_ansible_transport_blockers(*values: Any) -> list[str]:
+    combined = " ".join(str(value or "") for value in values)
+    blockers: list[str] = []
+    if "REPAIR_DENIED:invalid_command" in combined:
+        blockers.append(FORCED_COMMAND_BLOCKER)
+    return blockers
+
+
+def _playbook_roots(module_path: Path | None = None) -> list[Path]:
+    resolved_module_path = (module_path or Path(__file__)).resolve()
+    return [
+        Path("/app/infra/ansible"),
+        Path.cwd() / "infra" / "ansible",
+        *(parent / "infra" / "ansible" for parent in resolved_module_path.parents),
+    ]
+
+
+def _runtime_blockers(
+    *,
+    playbook_root: Path | None = None,
+    check_mode_ssh_key_path: Path | None = None,
+    check_mode_known_hosts_path: Path | None = None,
+) -> list[str]:
+    root = playbook_root or next((path for path in _playbook_roots() if path.exists()), None)
+    ssh_key_path = check_mode_ssh_key_path or _check_mode_ssh_key_path()
+    known_hosts_path = check_mode_known_hosts_path or _check_mode_known_hosts_path()
+    blockers: list[str] = []
+    if shutil.which("ansible-playbook") is None:
+        blockers.append("ansible_playbook_binary_missing")
+    if root is None:
+        blockers.append("ansible_playbook_catalog_missing")
+    elif not (root / "inventory" / "hosts.yml").exists():
+        blockers.append("ansible_inventory_missing")
+    if not ssh_key_path.is_file() or not os.access(ssh_key_path, os.R_OK):
+        blockers.append("ansible_check_mode_ssh_key_missing")
+    if not known_hosts_path.is_file() or not os.access(known_hosts_path, os.R_OK):
+        blockers.append("ansible_check_mode_known_hosts_missing")
+    return blockers
+
+
+def _safe_candidate(input_payload: dict[str, Any]) -> dict[str, Any]:
+    candidates = input_payload.get("executor_candidates")
+    if not isinstance(candidates, list) or not candidates:
+        raise ValueError("missing_executor_candidates")
+
+    for candidate in candidates:
+        if not isinstance(candidate, dict):
+            continue
+        catalog_id = str(candidate.get("catalog_id") or "")
+        catalog_item = get_ansible_catalog_item(catalog_id)
+        if not catalog_item:
+            continue
+        if catalog_item.get("supports_check_mode") is not True:
+            continue
+        if catalog_item.get("auto_apply_enabled") is True:
+            continue
+        catalog_playbook_path = str(catalog_item.get("playbook_path") or "")
+        candidate_playbook_path = str(candidate.get("playbook_path") or catalog_playbook_path)
+        check_mode_playbook_path = str(
+            candidate.get("check_mode_playbook_path")
+            or catalog_item.get("check_mode_playbook_path")
+            or catalog_playbook_path
+        )
+        if candidate_playbook_path not in {catalog_playbook_path, check_mode_playbook_path}:
+            continue
+        inventory_hosts = candidate.get("inventory_hosts") or catalog_item.get("inventory_hosts") or []
+        if (
+            isinstance(inventory_hosts, list)
+            and inventory_hosts
+            and all(isinstance(host, str) and _SAFE_HOST_RE.fullmatch(host) for host in inventory_hosts)
+        ):
+            return {
+                "catalog_id": catalog_id,
+                "playbook_path": check_mode_playbook_path,
+                "catalog_playbook_path": catalog_playbook_path,
+                "source_candidate_playbook_path": candidate_playbook_path,
+                "check_mode_playbook_path": check_mode_playbook_path,
+                "inventory_hosts": tuple(inventory_hosts),
+                "risk_level": str(candidate.get("risk_level") or catalog_item.get("risk_level") or ""),
+            }
+    raise ValueError("no_safe_check_mode_candidate")
+
+
+def build_ansible_check_mode_claim_input(
+    *,
+    source_candidate_op_id: str,
+    candidate_input: dict[str, Any],
+) -> dict[str, Any]:
+    safe = _safe_candidate(candidate_input)
+    incident_id = _incident_id_from_payload(candidate_input)
+    return {
+        "incident_id": incident_id,
+        "executor": "ansible",
+        "execution_backend": "ansible",
+        "execution_mode": "check_mode",
+        "transport_profile": settings.AWOOOP_ANSIBLE_CHECK_MODE_TRANSPORT_PROFILE,
+        "check_mode": True,
+        "diff": True,
+        "apply_enabled": False,
+        "approval_required_before_apply": True,
+        "source_candidate_op_id": source_candidate_op_id,
+        "catalog_id": safe["catalog_id"],
+        "playbook_path": safe["playbook_path"],
+        "catalog_playbook_path": safe["catalog_playbook_path"],
+        "source_candidate_playbook_path": safe["source_candidate_playbook_path"],
+        "check_mode_playbook_path": safe["check_mode_playbook_path"],
+        "inventory_hosts": list(safe["inventory_hosts"]),
+        "risk_level": safe["risk_level"],
+    }
+
+
+def _resolve_playbook_path(playbook_root: Path, playbook_path: str) -> Path:
+    relative = Path(playbook_path)
+    if relative.is_absolute() or not str(relative).startswith(str(_PLAYBOOK_PREFIX)):
+        raise ValueError("unsafe_playbook_path")
+    repo_root = playbook_root.parent.parent
+    resolved = (repo_root / relative).resolve()
+    allowed_root = (repo_root / _PLAYBOOK_PREFIX).resolve()
+    if allowed_root not in resolved.parents:
+        raise ValueError("playbook_outside_catalog")
+    if resolved.suffix not in {".yml", ".yaml"} or not resolved.exists():
+        raise ValueError("playbook_not_found")
+    return resolved
+
+
+def build_ansible_check_mode_command(
+    *,
+    playbook_path: str,
+    inventory_hosts: tuple[str, ...],
+    playbook_root: Path | None = None,
+    check_mode_ssh_key_path: Path | None = None,
+    check_mode_known_hosts_path: Path | None = None,
+) -> AnsibleCommandSpec:
+    root = playbook_root or next((path for path in _playbook_roots() if path.exists()), None)
+    if root is None:
+        raise ValueError("ansible_playbook_catalog_missing")
+    inventory_path = (root / "inventory" / "hosts.yml").resolve()
+    if not inventory_path.exists():
+        raise ValueError("ansible_inventory_missing")
+    if not inventory_hosts or not all(_SAFE_HOST_RE.fullmatch(host) for host in inventory_hosts):
+        raise ValueError("unsafe_inventory_hosts")
+
+    playbook_abs = _resolve_playbook_path(root, playbook_path)
+    ssh_key_path = check_mode_ssh_key_path or _check_mode_ssh_key_path()
+    known_hosts_path = check_mode_known_hosts_path or _check_mode_known_hosts_path()
+    ssh_common_args = (
+        f"-o UserKnownHostsFile={known_hosts_path} "
+        "-o IdentitiesOnly=yes -o BatchMode=yes"
+    )
+    extra_vars = {
+        "ansible_ssh_private_key_file": str(ssh_key_path),
+        "ansible_ssh_common_args": ssh_common_args,
+    }
+    command = [
+        "ansible-playbook",
+        "-i",
+        str(inventory_path),
+        str(playbook_abs),
+        "--check",
+        "--diff",
+        "--limit",
+        ",".join(inventory_hosts),
+        "--extra-vars",
+        json.dumps(extra_vars, ensure_ascii=False, separators=(",", ":")),
+    ]
+    env = {
+        **os.environ,
+        "ANSIBLE_HOST_KEY_CHECKING": "true",
+        "ANSIBLE_RETRY_FILES_ENABLED": "false",
+    }
+    return AnsibleCommandSpec(
+        command=command,
+        cwd=root,
+        env=env,
+        playbook_abs_path=playbook_abs,
+        inventory_abs_path=inventory_path,
+    )
+
+
+async def _run_ansible_command(spec: AnsibleCommandSpec, *, timeout_seconds: int) -> AnsibleRunResult:
+    started = time.monotonic()
+    process = await asyncio.create_subprocess_exec(
+        *spec.command,
+        cwd=str(spec.cwd),
+        env=spec.env,
+        stdout=asyncio.subprocess.PIPE,
+        stderr=asyncio.subprocess.PIPE,
+    )
+    timed_out = False
+    try:
+        stdout_bytes, stderr_bytes = await asyncio.wait_for(
+            process.communicate(),
+            timeout=timeout_seconds,
+        )
+    except TimeoutError:
+        timed_out = True
+        process.kill()
+        stdout_bytes, stderr_bytes = await process.communicate()
+    duration_ms = int((time.monotonic() - started) * 1000)
+    return AnsibleRunResult(
+        returncode=124 if timed_out else int(process.returncode or 0),
+        stdout=stdout_bytes.decode("utf-8", "replace"),
+        stderr=stderr_bytes.decode("utf-8", "replace"),
+        duration_ms=duration_ms,
+        timed_out=timed_out,
+    )
+
+
+def _build_result_payload(result: AnsibleRunResult) -> tuple[str, dict[str, Any], dict[str, Any], str | None]:
+    status = "success" if result.returncode == 0 else "failed"
+    stdout_tail = _tail(result.stdout, _STDOUT_LIMIT)
+    stderr_tail = _tail(result.stderr, _STDERR_LIMIT)
+    output = {
+        "executor": "ansible",
+        "execution_mode": "check_mode",
+        "transport_profile": settings.AWOOOP_ANSIBLE_CHECK_MODE_TRANSPORT_PROFILE,
+        "ssh_key_path": settings.AWOOOP_ANSIBLE_CHECK_MODE_SSH_KEY_PATH,
+        "known_hosts_path": settings.AWOOOP_ANSIBLE_CHECK_MODE_KNOWN_HOSTS_PATH,
+        "check_mode": True,
+        "apply_enabled": False,
+        "approval_required_before_apply": True,
+        "returncode": result.returncode,
+        "timed_out": result.timed_out,
+        "stdout_tail": stdout_tail,
+        "stderr_tail": stderr_tail,
+        "next_required_step": "approval_required_before_ansible_apply",
+    }
+    dry_run_result = {
+        "check_mode_executed": True,
+        "apply_executed": False,
+        "safe_to_apply_without_approval": False,
+        "transport_profile": settings.AWOOOP_ANSIBLE_CHECK_MODE_TRANSPORT_PROFILE,
+        "ssh_key_path": settings.AWOOOP_ANSIBLE_CHECK_MODE_SSH_KEY_PATH,
+        "known_hosts_path": settings.AWOOOP_ANSIBLE_CHECK_MODE_KNOWN_HOSTS_PATH,
+        "returncode": result.returncode,
+        "timed_out": result.timed_out,
+        "stdout_tail": stdout_tail,
+        "stderr_tail": stderr_tail,
+    }
+    error = None if result.returncode == 0 else (stderr_tail or f"ansible_check_mode_failed_rc_{result.returncode}")
+    return status, output, dry_run_result, error
+
+
+async def claim_pending_check_modes(
+    *,
+    project_id: str = "awoooi",
+    limit: int = 1,
+    candidate_max_age_hours: int | None = None,
+) -> list[AnsibleCheckModeClaim]:
+    """Claim pending Ansible candidates by inserting pending check-mode rows."""
+
+    claims: list[AnsibleCheckModeClaim] = []
+    max_age_hours = candidate_max_age_hours or settings.AWOOOP_ANSIBLE_CHECK_MODE_CANDIDATE_MAX_AGE_HOURS
+    async with get_db_context(project_id) as db:
+        result = await db.execute(
+            text("""
+                SELECT
+                    candidate.op_id,
+                    candidate.input
+                FROM automation_operation_log candidate
+                WHERE candidate.operation_type = 'ansible_candidate_matched'
+                  AND candidate.status = 'dry_run'
+                  AND candidate.input ->> 'executor' = 'ansible'
+                  AND candidate.created_at >= NOW() - (:candidate_max_age_hours * INTERVAL '1 hour')
+                  AND COALESCE((candidate.dry_run_result ->> 'check_mode_executed')::boolean, false) = false
+                  AND NOT EXISTS (
+                      SELECT 1
+                      FROM automation_operation_log existing
+                      WHERE existing.parent_op_id = candidate.op_id
+                        AND existing.operation_type IN (
+                            'ansible_check_mode_executed',
+                            'ansible_execution_skipped'
+                        )
+                  )
+                ORDER BY candidate.created_at ASC
+                LIMIT :limit
+                FOR UPDATE SKIP LOCKED
+            """),
+            {
+                "limit": max(1, limit),
+                "candidate_max_age_hours": max(1, max_age_hours),
+            },
+        )
+        rows = result.mappings().all()
+        for row in rows:
+            source_op_id = str(row["op_id"])
+            candidate_input = _json_loads(row["input"])
+            try:
+                claim_input = build_ansible_check_mode_claim_input(
+                    source_candidate_op_id=source_op_id,
+                    candidate_input=candidate_input,
+                )
+            except ValueError as exc:
+                await _insert_skipped_candidate(
+                    db,
+                    source_candidate_op_id=source_op_id,
+                    candidate_input=candidate_input,
+                    reason=str(exc),
+                )
+                continue
+            inserted = await db.execute(
+                text("""
+                    INSERT INTO automation_operation_log (
+                        operation_type, actor, status, incident_id,
+                        input, output, dry_run_result,
+                        parent_op_id, tags
+                    ) VALUES (
+                        'ansible_check_mode_executed',
+                        'ansible_check_mode_worker',
+                        'pending',
+                        :incident_db_id,
+                        CAST(:input AS jsonb),
+                        '{}'::jsonb,
+                        CAST(:dry_run_result AS jsonb),
+                        CAST(:parent_op_id AS uuid),
+                        :tags
+                    )
+                    RETURNING op_id
+                """),
+                {
+                    "incident_id": _incident_id_from_payload(claim_input),
+                    "incident_db_id": _automation_operation_log_incident_id(
+                        _incident_id_from_payload(claim_input)
+                    ),
+                    "input": json.dumps(claim_input, ensure_ascii=False),
+                    "dry_run_result": json.dumps({
+                        "check_mode_executed": False,
+                        "apply_executed": False,
+                        "claim_state": "claimed",
+                    }, ensure_ascii=False),
+                    "parent_op_id": source_op_id,
+                    "tags": ["ansible", "check_mode", "pending", "apply_locked"],
+                },
+            )
+            op_id = str(inserted.scalar_one())
+            claims.append(
+                AnsibleCheckModeClaim(
+                    op_id=op_id,
+                    source_candidate_op_id=source_op_id,
+                    incident_id=str(claim_input.get("incident_id") or ""),
+                    catalog_id=str(claim_input["catalog_id"]),
+                    playbook_path=str(claim_input["playbook_path"]),
+                    inventory_hosts=tuple(str(host) for host in claim_input["inventory_hosts"]),
+                    input_payload=claim_input,
+                )
+            )
+    return claims
+
+
+async def recent_ansible_transport_blockers(
+    *,
+    project_id: str = "awoooi",
+    cooldown_seconds: int | None = None,
+    include_repair_forced_command_blocker: bool | None = None,
+) -> list[str]:
+    """Return transport blockers observed from recent failed check-mode rows."""
+
+    cooldown = cooldown_seconds or settings.AWOOOP_ANSIBLE_CHECK_MODE_TRANSPORT_COOLDOWN_SECONDS
+    include_forced_blocker = (
+        _uses_repair_forced_command_transport()
+        if include_repair_forced_command_blocker is None
+        else include_repair_forced_command_blocker
+    )
+    async with get_db_context(project_id) as db:
+        result = await db.execute(
+            text("""
+                SELECT
+                    coalesce(output::text, '') AS output_text,
+                    coalesce(dry_run_result::text, '') AS dry_run_text,
+                    coalesce(error, '') AS error_text,
+                    coalesce(stderr_feed_back, '') AS stderr_text
+                FROM automation_operation_log
+                WHERE operation_type = 'ansible_check_mode_executed'
+                  AND status = 'failed'
+                  AND created_at >= NOW() - (:cooldown_seconds * INTERVAL '1 second')
+                ORDER BY created_at DESC
+                LIMIT 20
+            """),
+            {"cooldown_seconds": max(60, cooldown)},
+        )
+        blockers: set[str] = set()
+        for row in result.mappings().all():
+            detected = detect_ansible_transport_blockers(
+                row.get("output_text"),
+                row.get("dry_run_text"),
+                row.get("error_text"),
+                row.get("stderr_text"),
+            )
+            blockers.update(
+                blocker
+                for blocker in detected
+                if blocker != FORCED_COMMAND_BLOCKER or include_forced_blocker
+            )
+    return sorted(blockers)
+
+
+async def _insert_skipped_candidate(
+    db: Any,
+    *,
+    source_candidate_op_id: str,
+    candidate_input: dict[str, Any],
+    reason: str,
+) -> None:
+    input_payload = {
+        "incident_id": _incident_id_from_payload(candidate_input),
+        "executor": "ansible",
+        "execution_backend": "ansible",
+        "execution_mode": "check_mode",
+        "transport_profile": settings.AWOOOP_ANSIBLE_CHECK_MODE_TRANSPORT_PROFILE,
+        "check_mode": True,
+        "apply_enabled": False,
+        "source_candidate_op_id": source_candidate_op_id,
+        "not_used_reason": reason,
+    }
+    await db.execute(
+        text("""
+            INSERT INTO automation_operation_log (
+                operation_type, actor, status, incident_id,
+                input, output, dry_run_result,
+                parent_op_id, tags
+            ) VALUES (
+                'ansible_execution_skipped',
+                'ansible_check_mode_worker',
+                'dry_run',
+                :incident_db_id,
+                CAST(:input AS jsonb),
+                CAST(:output AS jsonb),
+                CAST(:dry_run_result AS jsonb),
+                CAST(:parent_op_id AS uuid),
+                :tags
+            )
+        """),
+        {
+            "incident_id": _incident_id_from_payload(input_payload),
+            "incident_db_id": _automation_operation_log_incident_id(
+                _incident_id_from_payload(input_payload)
+            ),
+            "input": json.dumps(input_payload, ensure_ascii=False),
+            "output": json.dumps({
+                "not_used_reason": reason,
+                "decision_effect": "skipped_before_runtime",
+            }, ensure_ascii=False),
+            "dry_run_result": json.dumps({
+                "check_mode_executed": False,
+                "apply_executed": False,
+                "skipped": True,
+                "reason": reason,
+            }, ensure_ascii=False),
+            "parent_op_id": source_candidate_op_id,
+            "tags": ["ansible", "check_mode", "skipped", "apply_locked"],
+        },
+    )
+
+
+async def finalize_check_mode_claim(
+    claim: AnsibleCheckModeClaim,
+    result: AnsibleRunResult,
+    *,
+    project_id: str = "awoooi",
+) -> None:
+    status, output, dry_run_result, error = _build_result_payload(result)
+    async with get_db_context(project_id) as db:
+        await db.execute(
+            text("""
+                UPDATE automation_operation_log
+                SET status = :status,
+                    output = CAST(:output AS jsonb),
+                    dry_run_result = CAST(:dry_run_result AS jsonb),
+                    error = :error,
+                    duration_ms = :duration_ms,
+                    stderr_feed_back = :stderr
+                WHERE op_id = CAST(:op_id AS uuid)
+            """),
+            {
+                "status": status,
+                "output": json.dumps(output, ensure_ascii=False),
+                "dry_run_result": json.dumps(dry_run_result, ensure_ascii=False),
+                "error": _tail(error or "", 2000) or None,
+                "duration_ms": result.duration_ms,
+                "stderr": _tail(result.stderr, _STDERR_LIMIT),
+                "op_id": claim.op_id,
+            },
+        )
+
+
+async def run_claimed_check_mode(
+    claim: AnsibleCheckModeClaim,
+    *,
+    timeout_seconds: int,
+    project_id: str = "awoooi",
+) -> AnsibleRunResult:
+    try:
+        spec = build_ansible_check_mode_command(
+            playbook_path=claim.playbook_path,
+            inventory_hosts=claim.inventory_hosts,
+        )
+        result = await _run_ansible_command(spec, timeout_seconds=timeout_seconds)
+    except Exception as exc:
+        result = AnsibleRunResult(
+            returncode=1,
+            stdout="",
+            stderr=f"ansible_check_mode_runtime_error: {exc}",
+            duration_ms=0,
+        )
+    await finalize_check_mode_claim(claim, result, project_id=project_id)
+    logger.info(
+        "ansible_check_mode_claim_completed",
+        op_id=claim.op_id,
+        source_candidate_op_id=claim.source_candidate_op_id,
+        incident_id=claim.incident_id,
+        catalog_id=claim.catalog_id,
+        returncode=result.returncode,
+        timed_out=result.timed_out,
+    )
+    return result
+
+
+async def run_pending_check_modes_once(
+    *,
+    project_id: str = "awoooi",
+    limit: int = 1,
+    timeout_seconds: int | None = None,
+) -> dict[str, Any]:
+    blockers = _runtime_blockers()
+    if blockers:
+        logger.warning("ansible_check_mode_runtime_blocked", blockers=blockers)
+        return {"claimed": 0, "completed": 0, "failed": 0, "blockers": blockers}
+    transport_blockers = await recent_ansible_transport_blockers(project_id=project_id)
+    if transport_blockers:
+        logger.warning("ansible_check_mode_transport_blocked", blockers=transport_blockers)
+        return {"claimed": 0, "completed": 0, "failed": 0, "blockers": transport_blockers}
+
+    claims = await claim_pending_check_modes(
+        project_id=project_id,
+        limit=limit,
+        candidate_max_age_hours=settings.AWOOOP_ANSIBLE_CHECK_MODE_CANDIDATE_MAX_AGE_HOURS,
+    )
+    completed = 0
+    failed = 0
+    for claim in claims:
+        result = await run_claimed_check_mode(
+            claim,
+            timeout_seconds=timeout_seconds or settings.AWOOOP_ANSIBLE_CHECK_MODE_TIMEOUT_SECONDS,
+            project_id=project_id,
+        )
+        completed += 1
+        if result.returncode != 0:
+            failed += 1
+    return {
+        "claimed": len(claims),
+        "completed": completed,
+        "failed": failed,
+        "blockers": [],
+    }

apps/api/src/services/awooop_deeplinks.py

@@ -0,0 +1,63 @@
+"""Shared AwoooP operator deeplinks used by Telegram-facing services."""
+
+from urllib.parse import quote
+
+AWOOOP_WEB_BASE_URL = "https://awoooi.wooo.work"
+
+
+def incident_runs_url(
+    incident_id: str,
+    *,
+    project_id: str = "awoooi",
+    locale: str = "zh-TW",
+) -> str:
+    safe_project_id = quote(str(project_id or "awoooi"), safe="")
+    safe_incident_id = quote(str(incident_id or ""), safe="")
+    return (
+        f"{AWOOOP_WEB_BASE_URL}/{locale}/awooop/runs"
+        f"?project_id={safe_project_id}&incident_id={safe_incident_id}"
+    )
+
+
+def incident_alerts_url(
+    incident_id: str,
+    *,
+    project_id: str = "awoooi",
+    locale: str = "zh-TW",
+) -> str:
+    safe_project_id = quote(str(project_id or "awoooi"), safe="")
+    safe_incident_id = quote(str(incident_id or ""), safe="")
+    return (
+        f"{AWOOOP_WEB_BASE_URL}/{locale}/alerts"
+        f"?project_id={safe_project_id}&incident_id={safe_incident_id}"
+    )
+
+
+def incident_truth_chain_button_row(
+    incident_id: str,
+    *,
+    project_id: str = "awoooi",
+) -> list[dict[str, str]]:
+    if not incident_id:
+        return []
+    return [
+        {
+            "text": "🔎 真相鏈",
+            "url": incident_alerts_url(incident_id, project_id=project_id),
+        },
+        {
+            "text": "🧭 Runs",
+            "url": incident_runs_url(incident_id, project_id=project_id),
+        },
+    ]
+
+
+def incident_truth_chain_reply_markup(
+    incident_id: str,
+    *,
+    project_id: str = "awoooi",
+) -> dict | None:
+    row = incident_truth_chain_button_row(incident_id, project_id=project_id)
+    if not row:
+        return None
+    return {"inline_keyboard": [row]}

apps/api/src/services/budget_service.py

@@ -336,7 +336,7 @@ async def _get_tenant_budget_limit(project_id: str) -> Decimal | None:
     try:
         from sqlalchemy import text
         from src.db.base import get_db_context
-        async with get_db_context() as db:
+        async with get_db_context(project_id) as db:
             row = await db.execute(
                 text("SELECT budget_limit_usd FROM awooop_projects WHERE project_id = :pid"),
                 {"pid": project_id},

apps/api/src/services/channel_hub.py

@@ -30,6 +30,7 @@ import asyncio
 import hashlib
 import html
 import json
+import re
 from datetime import UTC, datetime
 from typing import Any
 from uuid import NAMESPACE_URL, UUID, uuid5
@@ -39,13 +40,95 @@ from sqlalchemy import select, text
 from sqlalchemy.ext.asyncio import AsyncSession
 
 from src.db.awooop_models import AwoooPRunState
-from src.services.audit_sink import _redact_string
+from src.services.audit_sink import _redact_string, sanitize
 from src.services.platform_runtime import create_run
 
 logger = structlog.get_logger(__name__)
 
 # Progressive Feedback Policy：等待超過此秒數才發 interim 訊息
 _INTERIM_WAIT_SECONDS = 30
+_INBOUND_REDACTION_VERSION = "audit_sink_v1"
+_OUTBOUND_REDACTION_VERSION = "audit_sink_v1"
+_INCIDENT_ID_RE = re.compile(r"\bINC-\d{8}-[A-Z0-9]{6}\b")
+
+
+def _db_timestamp_now() -> datetime:
+    """Return UTC now in the timestamp shape accepted by the production DB path."""
+    return datetime.now(UTC).replace(tzinfo=None)
+
+
+def _compact_unique(values: list[str | None], *, limit: int = 20) -> list[str]:
+    """Return stable non-empty values without leaking duplicate source refs."""
+    return sorted({str(value).strip() for value in values if str(value or "").strip()})[:limit]
+
+
+def build_inbound_source_envelope(
+    *,
+    provider: str,
+    stage: str,
+    provider_event_id: str,
+    raw_event_id: str | None = None,
+    raw_content: str | None = None,
+    alertname: str | None = None,
+    severity: str | None = None,
+    namespace: str | None = None,
+    target_resource: str | None = None,
+    fingerprint: str | None = None,
+    incident_id: str | None = None,
+    approval_id: str | None = None,
+    source_url: str | None = None,
+    labels: dict[str, Any] | None = None,
+    annotations: dict[str, Any] | None = None,
+    extra: dict[str, Any] | None = None,
+) -> dict[str, Any]:
+    """Build a redaction-friendly inbound replay envelope for truth-chain use."""
+    content_sha256 = hashlib.sha256(raw_content.encode()).hexdigest() if raw_content else None
+    text_refs = _INCIDENT_ID_RE.findall(raw_content or "")
+    provider_name = str(provider or "unknown").strip().lower() or "unknown"
+    is_provider_heartbeat = str(stage or "").strip().lower() == "heartbeat"
+    source_refs = {
+        "event_ids": _compact_unique([raw_event_id]),
+        "incident_ids": _compact_unique([incident_id, *text_refs]),
+        "approval_ids": _compact_unique([approval_id]),
+        "alert_ids": _compact_unique([provider_event_id, raw_event_id]),
+        "fingerprints": _compact_unique([fingerprint]),
+        "sentry_issue_ids": _compact_unique(
+            [raw_event_id, provider_event_id]
+            if provider_name == "sentry" and not is_provider_heartbeat
+            else []
+        ),
+        "signoz_alerts": _compact_unique(
+            [raw_event_id, alertname]
+            if provider_name == "signoz" and not is_provider_heartbeat
+            else []
+        ),
+    }
+    envelope: dict[str, Any] = {
+        "schema_version": "inbound_source_envelope_v1",
+        "redaction_version": _INBOUND_REDACTION_VERSION,
+        "adapter": f"{provider_name}_webhook",
+        "provider": provider_name,
+        "stage": stage,
+        "provider_event_id": provider_event_id,
+        "source_url": source_url,
+        "content_sha256": content_sha256,
+        "content_length": len(raw_content) if raw_content is not None else 0,
+        "source_refs": source_refs,
+        "log_correlation": {
+            "alertname": alertname,
+            "severity": severity,
+            "namespace": namespace,
+            "target_resource": target_resource,
+            "fingerprint": fingerprint,
+        },
+        "labels": labels or {},
+        "annotations": annotations or {},
+    }
+    if extra:
+        envelope["extra"] = extra
+    sanitized = sanitize(envelope)
+    sanitized["content_sha256"] = content_sha256
+    return sanitized
 
 
 def _input_sha256(input_payload: dict[str, Any] | None) -> str | None:
@@ -82,11 +165,15 @@ async def ensure_completed_shadow_run(
             INSERT INTO awooop_run_state (
                 run_id, project_id, agent_id, state,
                 trigger_type, trigger_ref, is_shadow,
-                input_sha256, created_at, completed_at, timeout_at
+                input_sha256,
+                attempt_count, max_attempts, cost_usd, step_count,
+                created_at, completed_at, timeout_at
             ) VALUES (
                 :run_id, :project_id, :agent_id, 'completed',
                 :trigger_type, :trigger_ref, TRUE,
-                :input_sha256, NOW(), NOW(), NOW()
+                :input_sha256,
+                0, 3, 0.0000, 0,
+                NOW(), NOW(), NOW()
             )
             ON CONFLICT (run_id) DO NOTHING
             RETURNING run_id
@@ -117,6 +204,32 @@ def build_grouped_alert_run_id(project_id: str, provider_event_id: str) -> UUID:
     return uuid5(NAMESPACE_URL, f"awooop:grouped-alert:{project_id}:{provider_event_id}")
 
 
+def build_alertmanager_provider_event_id(alert_id: str, fingerprint: str, stage: str) -> str:
+    """建立 Alertmanager inbound event 的冪等 provider_event_id。"""
+    safe_alert_id = str(alert_id).strip() or "unknown"
+    safe_fingerprint = str(fingerprint).strip()[:32] or "no-fingerprint"
+    safe_stage = str(stage).strip()[:32] or "received"
+    return f"alertmanager:{safe_stage}:{safe_alert_id}:{safe_fingerprint}"
+
+
+def build_alertmanager_run_id(project_id: str, provider_event_id: str) -> UUID:
+    """為 Alertmanager inbound mirror 建立穩定 shadow run_id。"""
+    return uuid5(NAMESPACE_URL, f"awooop:alertmanager:{project_id}:{provider_event_id}")
+
+
+def build_external_alert_provider_event_id(provider: str, event_id: str, stage: str) -> str:
+    """建立 Sentry/SignOz 等外部告警 inbound event 的冪等 provider_event_id。"""
+    safe_provider = str(provider).strip().lower()[:32] or "external"
+    safe_event_id = str(event_id).strip()[:96] or "unknown"
+    safe_stage = str(stage).strip()[:32] or "received"
+    return f"{safe_provider}:{safe_stage}:{safe_event_id}"
+
+
+def build_external_alert_run_id(project_id: str, provider_event_id: str) -> UUID:
+    """為外部告警 inbound mirror 建立穩定 shadow run_id。"""
+    return uuid5(NAMESPACE_URL, f"awooop:external-alert:{project_id}:{provider_event_id}")
+
+
 # ─────────────────────────────────────────────────────────────────────────────
 # 入站事件記錄
 # ─────────────────────────────────────────────────────────────────────────────
@@ -132,6 +245,7 @@ async def mirror_inbound_event(
     channel_chat_id: str | None = None,
     content_type: str = "text",
     raw_content: str | None = None,
+    source_envelope: dict[str, Any] | None = None,
     attachment_sha256: str | None = None,
     provider_ts: datetime | None = None,
     run_id: UUID | None = None,
@@ -145,12 +259,32 @@ async def mirror_inbound_event(
     """
     content_hash: str | None = None
     content_preview: str | None = None
+    content_redacted: str | None = None
 
     if raw_content is not None:
         content_hash = hashlib.sha256(raw_content.encode()).hexdigest()
         # preview：redact 後截取前 256 字元
-        redacted = _redact_string(raw_content)
-        content_preview = redacted[:256] if len(redacted) > 256 else redacted
+        content_redacted = _redact_string(raw_content)
+        content_preview = (
+            content_redacted[:256] if len(content_redacted) > 256 else content_redacted
+        )
+
+    if source_envelope and source_envelope.get("schema_version") == "inbound_source_envelope_v1":
+        original_content_sha256 = source_envelope.get("content_sha256")
+        envelope = sanitize(source_envelope)
+        envelope.setdefault("redaction_version", _INBOUND_REDACTION_VERSION)
+        envelope["content_sha256"] = content_hash or original_content_sha256
+        envelope.setdefault("content_length", len(raw_content) if raw_content is not None else 0)
+    else:
+        envelope = build_inbound_source_envelope(
+            provider=channel_type,
+            stage="received",
+            provider_event_id=provider_event_id,
+            raw_event_id=provider_event_id,
+            raw_content=raw_content,
+            extra=source_envelope,
+        )
+    source_envelope_json = json.dumps(envelope, ensure_ascii=False, default=str)
 
     result = await db.execute(
         text("""
@@ -158,16 +292,28 @@ async def mirror_inbound_event(
                 project_id, channel_type, provider_event_id,
                 platform_subject_id, channel_user_id, channel_chat_id,
                 run_id, content_type, content_hash, content_preview,
+                content_redacted, redaction_version, source_envelope,
                 attachment_sha256, is_duplicate, provider_ts, received_at
             ) VALUES (
                 :project_id, :channel_type, :provider_event_id,
                 :platform_subject_id, :channel_user_id, :channel_chat_id,
                 :run_id, :content_type, :content_hash, :content_preview,
+                :content_redacted, :redaction_version, CAST(:source_envelope AS jsonb),
                 :attachment_sha256, :is_duplicate, :provider_ts, NOW()
             )
             ON CONFLICT (project_id, channel_type, provider_event_id) DO UPDATE SET
                 is_duplicate = TRUE,
-                run_id = COALESCE(EXCLUDED.run_id, awooop_conversation_event.run_id)
+                run_id = COALESCE(EXCLUDED.run_id, awooop_conversation_event.run_id),
+                content_redacted = COALESCE(
+                    awooop_conversation_event.content_redacted,
+                    EXCLUDED.content_redacted
+                ),
+                redaction_version = EXCLUDED.redaction_version,
+                source_envelope = CASE
+                    WHEN awooop_conversation_event.source_envelope = '{}'::jsonb
+                    THEN EXCLUDED.source_envelope
+                    ELSE awooop_conversation_event.source_envelope
+                END
             RETURNING event_id
         """),
         {
@@ -181,6 +327,9 @@ async def mirror_inbound_event(
             "content_type": content_type,
             "content_hash": content_hash,
             "content_preview": content_preview,
+            "content_redacted": content_redacted,
+            "redaction_version": _INBOUND_REDACTION_VERSION,
+            "source_envelope": source_envelope_json,
             "attachment_sha256": attachment_sha256,
             "is_duplicate": is_duplicate,
             "provider_ts": provider_ts,
@@ -205,6 +354,41 @@ def build_grouped_alert_provider_event_id(alert_id: str, fingerprint: str) -> st
     return f"alert-group:{safe_alert_id}:{safe_fingerprint}"
 
 
+def format_alertmanager_event_content(
+    *,
+    stage: str,
+    alert_id: str,
+    alertname: str,
+    severity: str,
+    namespace: str,
+    target_resource: str,
+    fingerprint: str,
+    notification_type: str | None = None,
+    alert_category: str | None = None,
+    incident_id: str | None = None,
+    approval_id: str | None = None,
+    repeat_count: int | None = None,
+) -> str:
+    """格式化 Alertmanager inbound mirror 摘要，讓 truth-chain 可回查。"""
+    head = f"Incident: {incident_id}" if incident_id else f"Fingerprint: {fingerprint}"
+    return "\n".join(
+        [
+            f"Alertmanager inbound {stage}",
+            head,
+            f"Alert ID: {alert_id}",
+            f"Approval: {approval_id or '-'}",
+            f"Alert: {alertname}",
+            f"Severity: {severity}",
+            f"Namespace: {namespace or 'default'}",
+            f"Target: {target_resource or '-'}",
+            f"Fingerprint: {fingerprint}",
+            f"Notification Type: {notification_type or '-'}",
+            f"Alert Category: {alert_category or '-'}",
+            f"Repeat Count: {repeat_count if repeat_count is not None else '-'}",
+        ]
+    )
+
+
 def format_grouped_alert_event_content(
     *,
     alert_id: str,
@@ -406,7 +590,7 @@ async def record_grouped_alert_event(
                 channel_chat_id=f"alert-group:{group_key}",
                 content_type="text",
                 raw_content=content,
-                provider_ts=datetime.now(UTC),
+                provider_ts=_db_timestamp_now(),
                 run_id=run_id,
             )
 
@@ -440,6 +624,267 @@ async def record_grouped_alert_event(
         return None
 
 
+async def record_alertmanager_event(
+    *,
+    project_id: str,
+    alert_id: str,
+    alertname: str,
+    severity: str,
+    namespace: str,
+    target_resource: str,
+    fingerprint: str,
+    stage: str,
+    notification_type: str | None = None,
+    alert_category: str | None = None,
+    incident_id: str | None = None,
+    approval_id: str | None = None,
+    repeat_count: int | None = None,
+    is_duplicate: bool = False,
+    source_url: str | None = None,
+    labels: dict[str, Any] | None = None,
+    annotations: dict[str, Any] | None = None,
+    source_extra: dict[str, Any] | None = None,
+) -> UUID | None:
+    """
+    將 Alertmanager inbound alert 鏡像到 AwoooP conversation_event。
+
+    Telegram 不應是唯一事實來源；每個 firing alert 至少要有 received
+    event，建立 incident/approval 後再補 incident_linked event 供 truth-chain
+    依 incident_id 回查。DB 失敗 fail-open，不影響 Alertmanager ACK。
+    """
+    try:
+        from src.db.base import get_db_context
+
+        incident_ref = str(incident_id) if incident_id else None
+        approval_ref = str(approval_id) if approval_id else None
+        provider_event_id = build_alertmanager_provider_event_id(
+            alert_id=alert_id,
+            fingerprint=fingerprint,
+            stage=stage,
+        )
+        content = format_alertmanager_event_content(
+            stage=stage,
+            alert_id=alert_id,
+            alertname=alertname,
+            severity=severity,
+            namespace=namespace,
+            target_resource=target_resource,
+            fingerprint=fingerprint,
+            notification_type=notification_type,
+            alert_category=alert_category,
+            incident_id=incident_ref,
+            approval_id=approval_ref,
+            repeat_count=repeat_count,
+        )
+        source_envelope = build_inbound_source_envelope(
+            provider="alertmanager",
+            stage=stage,
+            provider_event_id=provider_event_id,
+            raw_event_id=alert_id,
+            raw_content=content,
+            alertname=alertname,
+            severity=severity,
+            namespace=namespace,
+            target_resource=target_resource,
+            fingerprint=fingerprint,
+            incident_id=incident_ref,
+            approval_id=approval_ref,
+            source_url=source_url,
+            labels=labels,
+            annotations=annotations,
+            extra={
+                "notification_type": notification_type,
+                "alert_category": alert_category,
+                "repeat_count": repeat_count,
+                **(source_extra or {}),
+            },
+        )
+
+        async with get_db_context(project_id) as db:
+            run_id = build_alertmanager_run_id(project_id, provider_event_id)
+            await ensure_completed_shadow_run(
+                db,
+                project_id=project_id,
+                run_id=run_id,
+                agent_id="legacy-alertmanager-webhook",
+                trigger_type="alertmanager_inbound",
+                trigger_ref=provider_event_id,
+                input_payload={
+                    "stage": stage,
+                    "alert_id": alert_id,
+                    "alertname": alertname,
+                    "severity": severity,
+                    "namespace": namespace,
+                    "target_resource": target_resource,
+                    "fingerprint": fingerprint,
+                    "notification_type": notification_type,
+                    "alert_category": alert_category,
+                    "incident_id": incident_ref,
+                    "approval_id": approval_ref,
+                    "repeat_count": repeat_count,
+                },
+            )
+            event_id = await mirror_inbound_event(
+                db,
+                project_id=project_id,
+                channel_type="internal",
+                provider_event_id=provider_event_id,
+                platform_subject_id="alertmanager",
+                channel_user_id="alertmanager",
+                channel_chat_id=f"alertmanager:{namespace or 'default'}",
+                content_type="text",
+                raw_content=content,
+                source_envelope=source_envelope,
+                provider_ts=_db_timestamp_now(),
+                run_id=run_id,
+                is_duplicate=is_duplicate,
+            )
+
+        logger.info(
+            "alertmanager_event_recorded",
+            project_id=project_id,
+            alert_id=alert_id,
+            event_id=str(event_id),
+            stage=stage,
+            incident_id=incident_ref,
+            fingerprint=fingerprint,
+        )
+        return event_id
+    except Exception as exc:
+        logger.warning(
+            "alertmanager_event_record_failed",
+            project_id=project_id,
+            alert_id=alert_id,
+            stage=stage,
+            fingerprint=fingerprint,
+            error=str(exc),
+        )
+        return None
+
+
+async def record_external_alert_event(
+    *,
+    project_id: str,
+    provider: str,
+    event_id: str,
+    stage: str,
+    title: str,
+    severity: str,
+    namespace: str | None = None,
+    target_resource: str | None = None,
+    fingerprint: str | None = None,
+    incident_id: str | None = None,
+    approval_id: str | None = None,
+    source_url: str | None = None,
+    labels: dict[str, Any] | None = None,
+    annotations: dict[str, Any] | None = None,
+    payload: dict[str, Any] | None = None,
+    is_duplicate: bool = False,
+) -> UUID | None:
+    """
+    將 Sentry / SignOz 等非 Alertmanager 告警鏡像到 conversation_event。
+
+    這是 truth-chain 的最低共用入口：只寫 redacted content + source_envelope，
+    不改變原本 webhook 的通知、審批或自動化行為。
+    """
+    provider_name = str(provider or "external").strip().lower() or "external"
+    provider_event_id = build_external_alert_provider_event_id(provider_name, event_id, stage)
+    content = "\n".join([
+        f"{provider_name} inbound {stage}",
+        f"Event ID: {event_id}",
+        f"Title: {title}",
+        f"Severity: {severity}",
+        f"Namespace: {namespace or '-'}",
+        f"Target: {target_resource or '-'}",
+        f"Fingerprint: {fingerprint or '-'}",
+        f"Incident: {incident_id or '-'}",
+        f"Approval: {approval_id or '-'}",
+        f"Source URL: {source_url or '-'}",
+    ])
+    source_envelope = build_inbound_source_envelope(
+        provider=provider_name,
+        stage=stage,
+        provider_event_id=provider_event_id,
+        raw_event_id=event_id,
+        raw_content=content,
+        alertname=title,
+        severity=severity,
+        namespace=namespace,
+        target_resource=target_resource,
+        fingerprint=fingerprint,
+        incident_id=str(incident_id) if incident_id else None,
+        approval_id=str(approval_id) if approval_id else None,
+        source_url=source_url,
+        labels=labels,
+        annotations=annotations,
+        extra={
+            "payload": payload or {},
+        },
+    )
+
+    try:
+        from src.db.base import get_db_context
+
+        async with get_db_context(project_id) as db:
+            run_id = build_external_alert_run_id(project_id, provider_event_id)
+            await ensure_completed_shadow_run(
+                db,
+                project_id=project_id,
+                run_id=run_id,
+                agent_id=f"legacy-{provider_name}-webhook",
+                trigger_type=f"{provider_name}_inbound",
+                trigger_ref=provider_event_id,
+                input_payload={
+                    "provider": provider_name,
+                    "event_id": event_id,
+                    "stage": stage,
+                    "severity": severity,
+                    "namespace": namespace,
+                    "target_resource": target_resource,
+                    "fingerprint": fingerprint,
+                    "incident_id": str(incident_id) if incident_id else None,
+                    "approval_id": str(approval_id) if approval_id else None,
+                },
+            )
+            event_uuid = await mirror_inbound_event(
+                db,
+                project_id=project_id,
+                channel_type="internal",
+                provider_event_id=provider_event_id,
+                platform_subject_id=provider_name,
+                channel_user_id=provider_name,
+                channel_chat_id=f"{provider_name}:{namespace or 'default'}",
+                content_type="text",
+                raw_content=content,
+                source_envelope=source_envelope,
+                provider_ts=_db_timestamp_now(),
+                run_id=run_id,
+                is_duplicate=is_duplicate,
+            )
+
+        logger.info(
+            "external_alert_event_recorded",
+            project_id=project_id,
+            provider=provider_name,
+            event_id=event_id,
+            stage=stage,
+            conversation_event_id=str(event_uuid),
+            incident_id=str(incident_id) if incident_id else None,
+            approval_id=str(approval_id) if approval_id else None,
+        )
+        return event_uuid
+    except Exception as exc:
+        logger.warning(
+            "external_alert_event_record_failed",
+            project_id=project_id,
+            provider=provider_name,
+            event_id=event_id,
+            stage=stage,
+            error=str(exc),
+        )
+        return None
+
+
 # ─────────────────────────────────────────────────────────────────────────────
 # 出站訊息記錄
 # ─────────────────────────────────────────────────────────────────────────────
@@ -453,6 +898,7 @@ async def record_outbound_message(
     channel_chat_id: str,
     message_type: str,           # 'interim' | 'final' | 'error' | 'approval_request'
     content: str | None = None,
+    source_envelope: dict[str, Any] | None = None,
     provider_message_id: str | None = None,
     send_status: str = "pending",
     conversation_event_id: UUID | None = None,
@@ -467,12 +913,27 @@ async def record_outbound_message(
     """
     content_hash: str | None = None
     content_preview: str | None = None
+    content_redacted: str | None = None
     if content is not None:
         content_hash = hashlib.sha256(content.encode()).hexdigest()
-        redacted = _redact_string(content)
-        content_preview = redacted[:256]
+        content_redacted = _redact_string(content)
+        content_preview = content_redacted[:256]
+
+    envelope: dict[str, Any] = sanitize(source_envelope or {})
+    envelope.update({
+        "schema_version": "outbound_source_envelope_v1",
+        "redaction_version": _OUTBOUND_REDACTION_VERSION,
+        "content_sha256": content_hash,
+        "content_length": len(content) if content is not None else 0,
+    })
+    source_envelope_json = json.dumps(envelope, ensure_ascii=False, default=str)
 
     actual_status = "shadow" if is_shadow else send_status
+    sent_at = (
+        _db_timestamp_now()
+        if actual_status == "sent"
+        else None
+    )
 
     await ensure_completed_shadow_run(
         db,
@@ -495,14 +956,18 @@ async def record_outbound_message(
             INSERT INTO awooop_outbound_message (
                 project_id, run_id, conversation_event_id,
                 channel_type, channel_chat_id, message_type,
-                content_hash, content_preview, provider_message_id,
-                send_status, queued_at,
+                content_hash, content_preview, content_redacted,
+                redaction_version, source_envelope,
+                provider_message_id,
+                send_status, queued_at, sent_at,
                 triggered_by_state, waiting_since
             ) VALUES (
                 :project_id, :run_id, :conversation_event_id,
                 :channel_type, :channel_chat_id, :message_type,
-                :content_hash, :content_preview, :provider_message_id,
-                :send_status, NOW(),
+                :content_hash, :content_preview, :content_redacted,
+                :redaction_version, CAST(:source_envelope AS jsonb),
+                :provider_message_id,
+                :send_status, NOW(), :sent_at,
                 :triggered_by_state, :waiting_since
             )
             RETURNING message_id
@@ -516,8 +981,12 @@ async def record_outbound_message(
             "message_type": message_type,
             "content_hash": content_hash,
             "content_preview": content_preview,
+            "content_redacted": content_redacted,
+            "redaction_version": _OUTBOUND_REDACTION_VERSION,
+            "source_envelope": source_envelope_json,
             "provider_message_id": provider_message_id,
             "send_status": actual_status,
+            "sent_at": sent_at,
             "triggered_by_state": triggered_by_state,
             "waiting_since": waiting_since,
         },

apps/api/src/services/chat_manager.py

@@ -25,7 +25,7 @@ import structlog
 from src.core.config import get_settings
 from src.repositories.incident_repository import get_incident_repository
 from src.repositories.k8s_repository import get_k8s_repository
-from src.services.ollama_endpoint_resolver import resolve_ollama_endpoint
+from src.services.ollama_endpoint_resolver import resolve_ollama_order
 from src.utils.timezone import now_taipei
 
 logger = structlog.get_logger(__name__)
@@ -96,42 +96,51 @@ class ChatManager:
         settings = get_settings()
 
         model = settings.OPENCLAW_DEFAULT_MODEL
-        ollama_url = resolve_ollama_endpoint("interactive")
-        try:
-            async with httpx.AsyncClient(timeout=40.0) as client:
-                resp = await client.post(
-                    f"{ollama_url}/api/chat",
-                    json={
-                        "model": model,
-                        "stream": False,
-                        "messages": [
-                            {"role": "system", "content": system_prompt},
-                            {"role": "user", "content": user_message},
-                        ],
-                        "options": {"num_predict": 900, "temperature": 0.2},
-                    },
-                )
-                resp.raise_for_status()
-                data = resp.json()
-                raw = data.get("message", {}).get("content", "").strip()
-                text = re.sub(r"<think>.*?</think>", "", raw, flags=re.DOTALL).strip() or raw
+        async with httpx.AsyncClient(timeout=40.0) as client:
+            for endpoint in resolve_ollama_order("interactive"):
+                if not endpoint.url:
+                    continue
+                try:
+                    resp = await client.post(
+                        f"{endpoint.url}/api/chat",
+                        json={
+                            "model": model,
+                            "stream": False,
+                            "messages": [
+                                {"role": "system", "content": system_prompt},
+                                {"role": "user", "content": user_message},
+                            ],
+                            "options": {"num_predict": 900, "temperature": 0.2},
+                        },
+                    )
+                    resp.raise_for_status()
+                    data = resp.json()
+                    raw = data.get("message", {}).get("content", "").strip()
+                    text = re.sub(r"<think>.*?</think>", "", raw, flags=re.DOTALL).strip() or raw
 
-                eval_count = data.get("eval_count", 0)
-                prompt_eval_count = data.get("prompt_eval_count", 0)
-                total_tokens = eval_count + prompt_eval_count
+                    eval_count = data.get("eval_count", 0)
+                    prompt_eval_count = data.get("prompt_eval_count", 0)
+                    total_tokens = eval_count + prompt_eval_count
 
-                logger.info(
-                    "openclaw_ollama_chat_usage",
-                    model=model,
-                    endpoint=ollama_url,
-                    prompt_tokens=prompt_eval_count,
-                    output_tokens=eval_count,
-                )
+                    logger.info(
+                        "openclaw_ollama_chat_usage",
+                        model=model,
+                        endpoint=endpoint.url,
+                        provider=endpoint.provider_name,
+                        prompt_tokens=prompt_eval_count,
+                        output_tokens=eval_count,
+                    )
 
-                return f"{text}\n\n<i>🦙 {model} | {total_tokens} tokens | 免費</i>"
-        except Exception as e:
-            logger.warning("openclaw_chat_failed", error=str(e))
-            return None
+                    return f"{text}\n\n<i>🦙 {model} | {total_tokens} tokens | 免費</i>"
+                except Exception as e:
+                    logger.warning(
+                        "openclaw_chat_endpoint_failed",
+                        provider=endpoint.provider_name,
+                        endpoint=endpoint.url,
+                        error=str(e),
+                    )
+        logger.warning("openclaw_chat_failed_all_endpoints", model=model)
+        return None
 
     async def _call_nemotron(self, system_prompt: str, user_message: str) -> str | None:
         """
@@ -146,43 +155,60 @@ class ChatManager:
         system_prompt = f"{NEMOCLAW_PERSONA}\n{system_prompt}"
 
         # 2026-05-05 Codex: ADR-110 interactive lane，由 resolver 管理 GCP-A/GCP-B/111 拓撲
-        OLLAMA_URL = resolve_ollama_endpoint("interactive")
         MODEL = "deepseek-r1:14b"
 
-        try:
-            async with httpx.AsyncClient(timeout=120.0) as client:
-                resp = await client.post(
-                    f"{OLLAMA_URL}/api/chat",
-                    json={
-                        "model": MODEL,
-                        "stream": False,
-                        "messages": [
-                            {"role": "system", "content": system_prompt},
-                            {"role": "user", "content": user_message},
-                        ],
-                        "options": {"num_predict": 1200},
-                    },
-                )
-                resp.raise_for_status()
-                data = resp.json()
-                raw = data.get("message", {}).get("content", "").strip()
+        async with httpx.AsyncClient(timeout=120.0) as client:
+            for endpoint in resolve_ollama_order("interactive"):
+                if not endpoint.url:
+                    continue
+                try:
+                    resp = await client.post(
+                        f"{endpoint.url}/api/chat",
+                        json={
+                            "model": MODEL,
+                            "stream": False,
+                            # Ollama 0.24 separates deepseek-r1 thinking from final text.
+                            # Chat callers expect message.content to contain the answer.
+                            "think": False,
+                            "messages": [
+                                {"role": "system", "content": system_prompt},
+                                {"role": "user", "content": user_message},
+                            ],
+                            "options": {"num_predict": 1200},
+                        },
+                    )
+                    resp.raise_for_status()
+                    data = resp.json()
+                    raw = data.get("message", {}).get("content", "").strip()
 
-                # 過濾 deepseek-r1 的 <think>...</think> 推理區塊
-                text = re.sub(r"<think>.*?</think>", "", raw, flags=re.DOTALL).strip()
-                if not text:
-                    text = raw  # 萬一全是 think block，直接回傳原文
+                    # 過濾 deepseek-r1 的 <think>...</think> 推理區塊
+                    text = re.sub(r"<think>.*?</think>", "", raw, flags=re.DOTALL).strip()
+                    if not text:
+                        text = raw  # 萬一全是 think block，直接回傳原文
 
-                eval_count = data.get("eval_count", 0)
-                prompt_eval_count = data.get("prompt_eval_count", 0)
-                total_tokens = eval_count + prompt_eval_count
+                    eval_count = data.get("eval_count", 0)
+                    prompt_eval_count = data.get("prompt_eval_count", 0)
+                    total_tokens = eval_count + prompt_eval_count
 
-                logger.info("nemotron_ollama_usage", model=MODEL,
-                            prompt_tokens=prompt_eval_count, output_tokens=eval_count)
+                    logger.info(
+                        "nemotron_ollama_usage",
+                        model=MODEL,
+                        provider=endpoint.provider_name,
+                        prompt_tokens=prompt_eval_count,
+                        output_tokens=eval_count,
+                    )
 
-                return f"{text}\n\n<i>🦙 {MODEL} | {total_tokens} tokens | 免費</i>"
-        except Exception as e:
-            logger.warning("nemotron_chat_failed", model=MODEL, error=str(e))
-            return None
+                    return f"{text}\n\n<i>🦙 {MODEL} | {total_tokens} tokens | 免費</i>"
+                except Exception as e:
+                    logger.warning(
+                        "nemotron_chat_endpoint_failed",
+                        model=MODEL,
+                        provider=endpoint.provider_name,
+                        endpoint=endpoint.url,
+                        error=str(e),
+                    )
+        logger.warning("nemotron_chat_failed_all_endpoints", model=MODEL)
+        return None
 
     async def generate_response(
         self,

apps/api/src/services/dashboard_metrics_service.py

@@ -0,0 +1,31 @@
+"""
+Dashboard Metrics Service
+=========================
+Small DB-backed counters used by the war-room dashboard.
+"""
+
+from sqlalchemy import text
+
+from src.core.logging import get_logger
+from src.db.base import get_db_context
+
+logger = get_logger("awoooi.dashboard_metrics")
+
+
+async def fetch_pending_approval_count() -> int:
+    """Read the live HITL backlog from approval_records."""
+    try:
+        async with get_db_context() as db:
+            result = await db.execute(
+                text(
+                    """
+                    SELECT count(*)
+                    FROM approval_records
+                    WHERE upper(status::text) = 'PENDING'
+                    """
+                )
+            )
+            return int(result.scalar_one() or 0)
+    except Exception as exc:
+        logger.warning("dashboard_pending_approval_count_failed", error=str(exc))
+        return 0

apps/api/src/services/decision_fusion.py

@@ -29,6 +29,7 @@ import httpx
 import structlog
 
 from src.core.config import get_settings
+from src.services.ollama_endpoint_resolver import resolve_ollama_order
 
 if TYPE_CHECKING:
     from src.models.incident import Incident
@@ -186,9 +187,41 @@ class DecisionFusionEngine:
         # settings 延遲讀取（避免測試環境初始化問題）
         self._settings = get_settings()
 
-    @property
-    def _ollama_url(self) -> str:
-        return getattr(self._settings, "OLLAMA_URL", "http://34.143.170.20:11434")  # 2026-05-03 ogt: ADR-110 GCP-A Primary
+    async def _call_ollama_generate(
+        self,
+        *,
+        prompt: str,
+        timeout_sec: float,
+        num_predict: int,
+    ) -> str:
+        """Call Ollama in the global order: GCP-A -> GCP-B -> 111."""
+        last_error: Exception | None = None
+        async with httpx.AsyncClient(
+            timeout=httpx.Timeout(timeout_sec, connect=5.0)
+        ) as client:
+            for endpoint in resolve_ollama_order("deep_rca"):
+                if not endpoint.url:
+                    continue
+                try:
+                    resp = await client.post(
+                        f"{endpoint.url}{_OLLAMA_GENERATE_PATH}",
+                        json={
+                            "model": "qwen3:8b",
+                            "prompt": prompt,
+                            "stream": False,
+                            "options": {"num_predict": num_predict, "temperature": 0.1},
+                        },
+                    )
+                    resp.raise_for_status()
+                    return resp.json().get("response", "").strip()
+                except Exception as exc:
+                    last_error = exc
+                    logger.debug(
+                        "decision_fusion_ollama_endpoint_failed",
+                        provider=endpoint.provider_name,
+                        error=str(exc),
+                    )
+        raise RuntimeError(str(last_error) if last_error else "no_ollama_endpoint")
 
     # =========================================================================
     # Public API
@@ -322,21 +355,12 @@ class DecisionFusionEngine:
         )
 
         try:
-            async with httpx.AsyncClient(
-                timeout=httpx.Timeout(_HERMES_TIMEOUT_SEC, connect=5.0)
-            ) as client:
-                resp = await client.post(
-                    f"{self._ollama_url}{_OLLAMA_GENERATE_PATH}",
-                    json={
-                        "model": "qwen3:8b",
-                        "prompt": prompt,
-                        "stream": False,
-                        "options": {"num_predict": 16, "temperature": 0.1},
-                    },
-                )
-                if resp.status_code == 200:
-                    text = resp.json().get("response", "").strip()
-                    return self._extract_float(text, default=0.5)
+            text = await self._call_ollama_generate(
+                prompt=prompt,
+                timeout_sec=_HERMES_TIMEOUT_SEC,
+                num_predict=16,
+            )
+            return self._extract_float(text, default=0.5)
         except Exception as exc:
             logger.debug("hermes_score_failed", error=str(exc))
 
@@ -451,20 +475,11 @@ class DecisionFusionEngine:
             "只回覆一個 0-1 的小數，不要解釋。"
         )
 
-        async with httpx.AsyncClient(
-            timeout=httpx.Timeout(_ELEPHANT_TIMEOUT_SEC, connect=5.0)
-        ) as client:
-            resp = await client.post(
-                f"{self._ollama_url}{_OLLAMA_GENERATE_PATH}",
-                json={
-                    "model": "qwen3:8b",
-                    "prompt": prompt,
-                    "stream": False,
-                    "options": {"num_predict": 32, "temperature": 0.1},
-                },
-            )
-            resp.raise_for_status()
-            raw_text = resp.json().get("response", "").strip()
+        raw_text = await self._call_ollama_generate(
+            prompt=prompt,
+            timeout_sec=_ELEPHANT_TIMEOUT_SEC,
+            num_predict=32,
+        )
 
         # 移除 deepseek/qwen3 <think> 標籤
         clean = re.sub(r"<think>.*?</think>", "", raw_text, flags=re.DOTALL).strip()