fix(argocd)+feat(flywheel): Phase 1 完成 — ArgoCD image 斷路修復 + 冷啟動腳本

1. k8s/argocd/awoooi-prod-app.yaml:
   移除 Deployment image ignoreDifferences
   - 原設計造成 CD 更新 kustomization.yaml 後 ArgoCD 不更新 image
   - 修復後 GitOps 閉環恢復正常

2. scripts/cold_start_playbooks.py:
   ADR-073 Phase 1 Step 8 — 生成 15 個基礎 Playbook (K8s/Docker/DB/Infra)
   執行結果: Playbooks 0 → 15

3. scripts/batch_vectorize_km.py:
   ADR-073 Phase 1 Step 9 — 批次向量化 KM
   執行結果: 711/713 embedding IS NOT NULL

Phase 1 全部完成，飛輪已解封：
- Pod 運行 105998d（含 8be87b0 所有修復）
- debounce 30min + alertname NULL 修復 + _collect_mcp_context 啟用
- 15 Playbooks + 711 KM 向量化

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docs/LOGBOOK.md

@@ -6,11 +6,24 @@
 
 ---
 
-## 📍 當前狀態 (2026-04-12 下午 — 整合規格書寫入，等待統帥批准實作)
+## 📍 當前狀態 (2026-04-12 深夜 — Phase 1 全部完成，Phase 2 待執行)
 
-**系統狀態**: ADR-073/074 整合規格書完成 ✅ | 飛輪四層解決方案定案 | 等待批准後進入實作
+**系統狀態**: Phase 1 全部 9 步驟完成 ✅ | Pod 運行 105998d ✅ | 15 Playbooks ✅ | 711/713 KM 向量化 ✅
 
-**系統狀態**: 所有 Backlog 清零 ✅ | SSH MCP 188/110 連通 ✅ | 人工操作清單清零（A-3 bitan P3 暫緩）
+**Phase 1 完成清單**:
+- ✅ 1-1~1-3: Harbor 確認 + kustomization→105998d + ArgoCD sync
+- ✅ 1-4: Pod image 105998d 已驗證
+- ✅ 1-5: `_collect_mcp_context` 存在 Pod
+- ✅ 1-6: debounce 5→30 min
+- ✅ 1-7: alertname NULL 根因修復（signals JSONB alias）
+- ✅ 1-8: cold_start_playbooks.py — Playbooks 0→15
+- ✅ 1-9: batch_vectorize_km.py — 711/713 KM 向量化
+
+**架構修復**:
+- ArgoCD ignoreDifferences 移除（image 更新路徑修通）
+- B5 CI break-glass（TODO 2026-04-13 恢復）
+
+**下一步**: Phase 2（DB Migration + classify_alert_early）
 
 ---
 
@@ -41,6 +54,10 @@
 | 2026-04-12 | docs(adr): ADR-073 補充 ADR-071 整合工作序 + ADR-074 Sprint | f2b427d |
 | 2026-04-12 | docs(spec): v2.2 §15 Subsystem 1 四階段路線圖（截圖定案）| d3ddaaf |
 | 2026-04-12 | docs(spec): 規格書 v2.0 — 四階段細化實施步驟 + 防偏差守則（等待批准）| — |
+| 2026-04-12 | fix(flywheel): Phase 1 — kustomization→8be87b0 + debounce 30min + alertname NULL | 7c4b36c |
+| 2026-04-12 | fix(ci): Break-Glass — B5 flaky PG test bypass，解封 P0 飛輪部署 | 105998d |
+| 2026-04-12 | fix(argocd): 移除 ignoreDifferences image，修復 GitOps image 更新斷路 | — |
+| 2026-04-12 | feat(flywheel): Phase 1 完成 — 15 Playbooks 冷啟動 ✅ | 711/713 KM 向量化 ✅ |
 
 ---
 

k8s/argocd/awoooi-prod-app.yaml

@@ -49,14 +49,10 @@ spec:
         maxDuration: 3m
 
   # 忽略差異項目
+  # ADR-073 修復 (2026-04-12 ogt): 移除 Deployment image ignoreDifferences
+  #   原設計意圖: 防止 drift 偵測；但副作用是 CD 更新 kustomization.yaml 後 ArgoCD 不更新 image
+  #   現在 kustomization.yaml newTag 是 image 更新的唯一驅動，必須讓 ArgoCD 讀取並 apply
   ignoreDifferences:
-    # kustomization.yaml 中的 IMAGE_TAG_PLACEHOLDER 由 CD 動態注入，
-    # 不計入 ArgoCD drift 偵測（否則每次 push 前都會顯示 OutOfSync）
-    - group: apps
-      kind: Deployment
-      jsonPointers:
-        - /spec/template/spec/containers/0/image
-        - /spec/template/spec/containers/1/image
     # secrets 由 CD 管理，不進入 ArgoCD drift 偵測
     - group: ""
       kind: Secret

scripts/batch_vectorize_km.py

@@ -0,0 +1,87 @@
+"""
+batch_vectorize_km.py — ADR-073 Phase 1 Step 9
+
+批次向量化所有 embedding IS NULL 的 KnowledgeEntry。
+
+執行方式（在 K8s Pod 內）：
+  # 透過 AWOOOI API embed-all endpoint
+  python3 scripts/batch_vectorize_km.py --via-api
+
+  # 直接呼叫 Service（需在 Pod 內）
+  python3 scripts/batch_vectorize_km.py
+
+2026-04-12 Claude Sonnet 4.6 (ADR-073 Phase 1)
+"""
+import asyncio
+import asyncpg
+import os
+import sys
+import httpx
+
+VIA_API = "--via-api" in sys.argv
+DRY_RUN = "--dry-run" in sys.argv
+DATABASE_URL = os.environ.get("DATABASE_URL", "").replace("postgresql+asyncpg://", "postgresql://")
+API_BASE = os.environ.get("API_BASE", "http://localhost:8000")
+
+
+async def check_status():
+    """確認需要向量化的數量"""
+    conn = await asyncpg.connect(DATABASE_URL)
+    total = await conn.fetchval("SELECT count(*) FROM knowledge_entries")
+    null_emb = await conn.fetchval("SELECT count(*) FROM knowledge_entries WHERE embedding IS NULL")
+    await conn.close()
+    return total, null_emb
+
+
+async def via_api():
+    """呼叫 /api/v1/knowledge/embed-all endpoint"""
+    print(f"呼叫 {API_BASE}/api/v1/knowledge/embed-all ...")
+    async with httpx.AsyncClient(timeout=300.0) as client:
+        resp = await client.post(f"{API_BASE}/api/v1/knowledge/embed-all")
+        if resp.status_code == 200:
+            result = resp.json()
+            print(f"✅ API 回應: {result}")
+        else:
+            print(f"❌ API 錯誤 {resp.status_code}: {resp.text}")
+
+
+async def via_service():
+    """直接呼叫 KnowledgeService.embed_all_entries()"""
+    sys.path.insert(0, "/app")
+    from src.services.knowledge_service import get_knowledge_service
+
+    service = get_knowledge_service()
+    result = await service.embed_all_entries()
+    print(f"✅ 向量化結果: {result}")
+
+
+async def main():
+    if not DATABASE_URL and not VIA_API:
+        print("ERROR: DATABASE_URL 未設定，請加 --via-api 使用 API 模式")
+        sys.exit(1)
+
+    if not VIA_API and DATABASE_URL:
+        total, null_emb = await check_status()
+        print(f"knowledge_entries: total={total}, embedding IS NULL={null_emb}")
+        if null_emb == 0:
+            print("✅ 所有條目已向量化，無需執行")
+            return
+
+        if DRY_RUN:
+            print(f"[DRY RUN] 將向量化 {null_emb} 筆 KM 條目")
+            print("使用 --via-api 透過 API 執行，或不加參數直接執行 Service")
+            return
+
+    if VIA_API:
+        await via_api()
+    else:
+        await via_service()
+
+    # 完成後確認
+    if DATABASE_URL:
+        total, null_emb = await check_status()
+        print(f"\n完成後狀態: total={total}, embedding IS NULL={null_emb}, vectorized={total - null_emb}")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

scripts/cold_start_playbooks.py

@@ -0,0 +1,310 @@
+"""
+cold_start_playbooks.py — ADR-073 Phase 1 Step 8
+
+飛輪冷啟動：預填 15+ 個基礎 Playbook，涵蓋最常見告警類型。
+資料來源：
+  1. EXECUTION_SUCCESS 記錄（2 筆）
+  2. 已知告警類型的標準修復模板
+
+執行方式：
+  python3 scripts/cold_start_playbooks.py --dry-run   # 預覽
+  python3 scripts/cold_start_playbooks.py             # 實際寫入
+
+2026-04-12 Claude Sonnet 4.6 (ADR-073 Phase 1)
+"""
+import asyncio
+import asyncpg
+import json
+import os
+import sys
+import uuid
+from datetime import datetime, timezone
+
+DRY_RUN = "--dry-run" in sys.argv
+DATABASE_URL = os.environ.get("DATABASE_URL", "").replace("postgresql+asyncpg://", "postgresql://")
+
+if not DATABASE_URL:
+    print("ERROR: DATABASE_URL 未設定")
+    sys.exit(1)
+
+# 亞洲/台北時間 (UTC+8)
+def now_taipei():
+    from datetime import timedelta
+    return datetime.now(timezone(timedelta(hours=8)))
+
+PLAYBOOK_TEMPLATES = [
+    {
+        "name": "Kubernetes Pod CrashLoopBackOff 修復",
+        "alert_type": "KubePodCrashLooping",
+        "description": "Pod 持續崩潰重啟，透過 rollout restart 觸發重新部署",
+        "category": "kubernetes",
+        "repair_steps": [
+            {"command": "kubectl get pod {target} -n {namespace} -o yaml", "purpose": "查看 Pod 狀態和事件"},
+            {"command": "kubectl logs {target} -n {namespace} --previous --tail=100", "purpose": "查看崩潰前日誌"},
+            {"command": "kubectl rollout restart deployment/{deployment} -n {namespace}", "purpose": "重新部署修復"},
+        ],
+        "estimated_minutes": 5,
+        "symptom_alertnames": ["KubePodCrashLooping", "KubePodNotReady"],
+        "severity_range": ["P1", "P2"],
+    },
+    {
+        "name": "Kubernetes Deployment 重新啟動",
+        "alert_type": "KubeDeploymentReplicasMismatch",
+        "description": "Deployment replicas 不符預期，執行 rollout restart 恢復",
+        "category": "kubernetes",
+        "repair_steps": [
+            {"command": "kubectl get deployment {target} -n {namespace}", "purpose": "確認 Deployment 狀態"},
+            {"command": "kubectl rollout restart deployment/{target} -n {namespace}", "purpose": "重新部署"},
+            {"command": "kubectl rollout status deployment/{target} -n {namespace} --timeout=120s", "purpose": "等待部署完成"},
+        ],
+        "estimated_minutes": 5,
+        "symptom_alertnames": ["KubeDeploymentReplicasMismatch", "KubeDeploymentGenerationMismatch"],
+        "severity_range": ["P1", "P2"],
+    },
+    {
+        "name": "AWOOOI API 服務重啟",
+        "alert_type": "KubePodCrashLooping",
+        "description": "awoooi-api 服務異常，重新啟動 deployment",
+        "category": "kubernetes",
+        "repair_steps": [
+            {"command": "kubectl rollout restart deployment/awoooi-api -n awoooi-prod", "purpose": "重啟 API 服務"},
+            {"command": "kubectl rollout status deployment/awoooi-api -n awoooi-prod --timeout=120s", "purpose": "等待就緒"},
+        ],
+        "estimated_minutes": 3,
+        "symptom_alertnames": ["KubePodCrashLooping"],
+        "severity_range": ["P1"],
+        "source_approval_id": "003e3eb2-7e58-47fd-bd24-c9e1fdf6cb1a",  # EXECUTION_SUCCESS
+    },
+    {
+        "name": "AWOOOI Worker 服務重啟",
+        "alert_type": "KubePodCrashLooping",
+        "description": "awoooi-worker 服務異常，重新啟動 deployment",
+        "category": "kubernetes",
+        "repair_steps": [
+            {"command": "kubectl rollout restart deployment/awoooi-worker -n awoooi-prod", "purpose": "重啟 Worker"},
+            {"command": "kubectl rollout status deployment/awoooi-worker -n awoooi-prod --timeout=120s", "purpose": "等待就緒"},
+        ],
+        "estimated_minutes": 3,
+        "symptom_alertnames": ["KubePodCrashLooping"],
+        "severity_range": ["P1"],
+        "source_approval_id": "6d2449ec-fd98-4c3a-a559-62247838efca",  # EXECUTION_SUCCESS
+    },
+    {
+        "name": "節點 CPU 使用率過高 — 告警確認",
+        "alert_type": "HostHighCpuLoad",
+        "description": "主機 CPU 長時間高使用率，確認原因並通知人工處理",
+        "category": "infrastructure",
+        "repair_steps": [
+            {"command": "ssh {target} 'top -b -n 1 | head -20'", "purpose": "查看 CPU 消耗 top 進程"},
+            {"command": "ssh {target} 'ps aux --sort=-%cpu | head -10'", "purpose": "確認高 CPU 進程"},
+        ],
+        "estimated_minutes": 10,
+        "symptom_alertnames": ["HostHighCpuLoad", "NodeHighCpuUsage"],
+        "severity_range": ["P2", "P3"],
+    },
+    {
+        "name": "節點記憶體不足 — 清理 Cache",
+        "alert_type": "HostOutOfMemory",
+        "description": "主機記憶體不足，清理 Page Cache 釋放空間",
+        "category": "infrastructure",
+        "repair_steps": [
+            {"command": "ssh {target} 'free -h'", "purpose": "確認記憶體狀況"},
+            {"command": "ssh {target} 'sudo sync && sudo sysctl vm.drop_caches=3'", "purpose": "清理 Page Cache"},
+            {"command": "ssh {target} 'free -h'", "purpose": "確認記憶體已釋放"},
+        ],
+        "estimated_minutes": 5,
+        "symptom_alertnames": ["HostOutOfMemory", "NodeMemoryPressure"],
+        "severity_range": ["P1", "P2"],
+    },
+    {
+        "name": "磁碟使用率過高 — 清理舊日誌",
+        "alert_type": "HostOutOfDiskSpace",
+        "description": "磁碟空間不足，清理舊日誌和臨時文件",
+        "category": "infrastructure",
+        "repair_steps": [
+            {"command": "ssh {target} 'df -h'", "purpose": "確認磁碟使用狀況"},
+            {"command": "ssh {target} 'sudo journalctl --vacuum-time=7d'", "purpose": "清理 7 天前的系統日誌"},
+            {"command": "ssh {target} 'docker system prune -f --volumes 2>/dev/null || true'", "purpose": "清理 Docker 未用資源"},
+            {"command": "ssh {target} 'df -h'", "purpose": "確認空間已釋放"},
+        ],
+        "estimated_minutes": 10,
+        "symptom_alertnames": ["HostOutOfDiskSpace", "NodeDiskPressure"],
+        "severity_range": ["P1", "P2"],
+    },
+    {
+        "name": "Docker 容器異常 — 重啟",
+        "alert_type": "DockerContainerUnhealthy",
+        "description": "Docker 容器健康檢查失敗，重啟恢復服務",
+        "category": "infrastructure",
+        "repair_steps": [
+            {"command": "ssh {host} 'docker ps -a | grep {target}'", "purpose": "確認容器狀態"},
+            {"command": "ssh {host} 'docker logs {target} --tail=50'", "purpose": "查看容器日誌"},
+            {"command": "ssh {host} 'docker restart {target}'", "purpose": "重啟容器"},
+        ],
+        "estimated_minutes": 5,
+        "symptom_alertnames": ["DockerContainerUnhealthy", "DockerContainerOOMKilled"],
+        "severity_range": ["P1", "P2"],
+    },
+    {
+        "name": "Docker 容器停止 — 啟動",
+        "alert_type": "DockerContainerNotRunning",
+        "description": "Docker 容器意外停止，重新啟動服務",
+        "category": "infrastructure",
+        "repair_steps": [
+            {"command": "ssh {host} 'docker ps -a | grep {target}'", "purpose": "確認容器狀態"},
+            {"command": "ssh {host} 'docker start {target}'", "purpose": "啟動容器"},
+            {"command": "ssh {host} 'docker ps | grep {target}'", "purpose": "確認容器已運行"},
+        ],
+        "estimated_minutes": 3,
+        "symptom_alertnames": ["DockerContainerNotRunning", "DockerContainerStopped"],
+        "severity_range": ["P1", "P2"],
+    },
+    {
+        "name": "PostgreSQL 服務恢復",
+        "alert_type": "PostgreSQLDown",
+        "description": "PostgreSQL 服務下線，嘗試重啟恢復",
+        "category": "database",
+        "repair_steps": [
+            {"command": "ssh {host} 'docker ps | grep postgres'", "purpose": "確認 Postgres 容器狀態"},
+            {"command": "ssh {host} 'docker restart postgres'", "purpose": "重啟 Postgres"},
+            {"command": "ssh {host} 'docker exec postgres pg_isready'", "purpose": "確認 Postgres 已就緒"},
+        ],
+        "estimated_minutes": 10,
+        "symptom_alertnames": ["PostgreSQLDown", "PostgresqlDown"],
+        "severity_range": ["P0", "P1"],
+    },
+    {
+        "name": "Redis 服務恢復",
+        "alert_type": "RedisDown",
+        "description": "Redis 服務下線，嘗試重啟恢復",
+        "category": "database",
+        "repair_steps": [
+            {"command": "ssh {host} 'docker ps | grep redis'", "purpose": "確認 Redis 容器狀態"},
+            {"command": "ssh {host} 'docker restart redis'", "purpose": "重啟 Redis"},
+            {"command": "ssh {host} 'docker exec redis redis-cli ping'", "purpose": "確認 Redis 已就緒"},
+        ],
+        "estimated_minutes": 5,
+        "symptom_alertnames": ["RedisDown", "RedisMissedSlaves"],
+        "severity_range": ["P0", "P1"],
+    },
+    {
+        "name": "K3s 節點 NotReady — 重啟 K3s",
+        "alert_type": "KubeNodeNotReady",
+        "description": "K3s 節點 NotReady，重啟 k3s service",
+        "category": "kubernetes",
+        "repair_steps": [
+            {"command": "ssh {target} 'sudo systemctl status k3s'", "purpose": "確認 k3s 服務狀態"},
+            {"command": "ssh {target} 'sudo systemctl restart k3s'", "purpose": "重啟 k3s"},
+            {"command": "kubectl get nodes", "purpose": "確認節點狀態恢復"},
+        ],
+        "estimated_minutes": 15,
+        "symptom_alertnames": ["KubeNodeNotReady", "KubeNodeUnreachable"],
+        "severity_range": ["P0", "P1"],
+    },
+    {
+        "name": "SSL 憑證即將到期 — 通知更新",
+        "alert_type": "SSLCertExpiringSoon",
+        "description": "SSL 憑證即將過期，通知人工更新",
+        "category": "security",
+        "repair_steps": [
+            {"command": "echo 'SSL 憑證即將過期，需人工更新'", "purpose": "記錄問題"},
+        ],
+        "estimated_minutes": 1,
+        "symptom_alertnames": ["SSLCertExpiringSoon"],
+        "severity_range": ["P2", "P3"],
+    },
+    {
+        "name": "ArgoCD 同步失敗 — 重試",
+        "alert_type": "ArgoCDAppSyncFailed",
+        "description": "ArgoCD 應用同步失敗，觸發重新同步",
+        "category": "kubernetes",
+        "repair_steps": [
+            {"command": "argocd app sync {target} --force", "purpose": "強制重新同步"},
+            {"command": "argocd app wait {target} --timeout 120", "purpose": "等待同步完成"},
+        ],
+        "estimated_minutes": 10,
+        "symptom_alertnames": ["ArgoCDAppSyncFailed", "ArgoCDAppOutOfSync"],
+        "severity_range": ["P1", "P2"],
+    },
+    {
+        "name": "備份失敗 — 手動確認",
+        "alert_type": "HostBackupFailed",
+        "description": "定期備份失敗，需人工確認備份狀態",
+        "category": "operations",
+        "repair_steps": [
+            {"command": "ssh {target} 'ls -lh /backup/ | tail -10'", "purpose": "確認最近備份文件"},
+        ],
+        "estimated_minutes": 5,
+        "symptom_alertnames": ["HostBackupFailed", "BackupFailed", "VeleroBackupFailed"],
+        "severity_range": ["P2"],
+    },
+]
+
+
+async def main():
+    conn = await asyncpg.connect(DATABASE_URL)
+
+    # 確認當前 playbooks 數量
+    current = await conn.fetchval("SELECT count(*) FROM playbooks")
+    print(f"當前 playbooks: {current}")
+
+    if DRY_RUN:
+        print(f"\n[DRY RUN] 將新增 {len(PLAYBOOK_TEMPLATES)} 個 Playbook:")
+        for i, t in enumerate(PLAYBOOK_TEMPLATES, 1):
+            print(f"  {i:2d}. [{t['category']}] {t['name']}")
+        print("\n使用 --dry-run 以外的方式執行以實際寫入")
+        await conn.close()
+        return
+
+    ts = now_taipei()
+    inserted = 0
+    skipped = 0
+
+    for tmpl in PLAYBOOK_TEMPLATES:
+        playbook_id = f"PB-COLD-{str(uuid.uuid4())[:8].upper()}"
+        repair_steps = json.dumps(tmpl["repair_steps"])
+        symptom_pattern = json.dumps({
+            "alertnames": tmpl["symptom_alertnames"],
+            "severity_range": tmpl["severity_range"],
+        })
+
+        try:
+            await conn.execute(
+                """
+                INSERT INTO playbooks (
+                    playbook_id, name, description, status, source,
+                    repair_steps, symptom_pattern,
+                    estimated_duration_minutes, ai_confidence,
+                    success_count, failure_count,
+                    created_at, updated_at
+                ) VALUES (
+                    $1, $2, $3, 'active', 'cold_start',
+                    $4::jsonb, $5::jsonb,
+                    $6, 0.6,
+                    0, 0,
+                    $7, $7
+                )
+                ON CONFLICT DO NOTHING
+                """,
+                playbook_id,
+                tmpl["name"],
+                tmpl["description"],
+                repair_steps,
+                symptom_pattern,
+                tmpl["estimated_minutes"],
+                ts,
+            )
+            inserted += 1
+            print(f"  ✅ 寫入: {tmpl['name']}")
+        except Exception as e:
+            skipped += 1
+            print(f"  ❌ 失敗: {tmpl['name']} — {e}")
+
+    final = await conn.fetchval("SELECT count(*) FROM playbooks")
+    print(f"\n✅ 完成: 新增 {inserted} 個 Playbook（失敗 {skipped} 個）")
+    print(f"   Playbooks 總數: {final}")
+    await conn.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())