fix(governance): cast dispatch status filter

apps/api/src/services/governance_query_service.py

@@ -275,14 +275,14 @@ async def _query_dispatch_table(
             NULL::text AS operator_note
         FROM governance_remediation_dispatch d
         JOIN ai_governance_events e ON e.id = d.governance_event_id
-        WHERE d.dispatch_status = :dispatch_status
+        WHERE d.dispatch_status = CAST(:dispatch_status AS governance_dispatch_status)
         ORDER BY d.dispatched_at DESC
     """)
 
     count_sql = text("""
         SELECT count(*) AS cnt
         FROM governance_remediation_dispatch
-        WHERE dispatch_status = :dispatch_status
+        WHERE dispatch_status = CAST(:dispatch_status AS governance_dispatch_status)
     """)
 
     async with get_db_context() as db:

apps/api/tests/test_ai_governance_endpoints.py

@@ -288,6 +288,7 @@ class TestQueueEndpoint:
         assert "d.dispatched_at AS created_at" in source
         assert "ORDER BY d.dispatched_at DESC" in source
         assert "NULL::text AS operator_note" in source
+        assert "CAST(:dispatch_status AS governance_dispatch_status)" in source
         assert "d.created_at" not in source
         assert "d.operator_note" not in source
 

docs/LOGBOOK.md

@@ -5,11 +5,13 @@
 **根因**：
 - `governance/events`：production `ai_governance_events.details.remediation` 已有 dict 形態，例如 `{"items": [...]}`；read model `GovernanceEvent.remediation` 期待字串，Pydantic validation 造成 500。
 - `governance/queue`：查詢仍讀 `governance_remediation_dispatch.created_at` / `operator_note`，但 production migration schema 實際是 `dispatched_at` / `created_by`，沒有 `created_at` / `operator_note`。
+- `governance/queue`：第二層 production 差異是 `dispatch_status` 為 PostgreSQL enum；SQLAlchemy 參數被送成 varchar，未明確 cast 時 Postgres 會拒絕 enum = varchar 比較，導致真表被誤判成 `table_pending=true`。
 
 **修正**：
 - `governance_query_service._extract_remediation()` 將 `details.remediation` 的 string / dict / list 正規化成短字串，避免歷史治理事件破壞 response schema。
 - `_to_governance_event()` 對非 dict details 做 read-side guard。
 - `_query_dispatch_table()` 對齊 production schema：以 `d.dispatched_at AS created_at`、`NULL::text AS operator_note` 相容現有前端 DTO，不改 DB schema。
+- `_query_dispatch_table()` 對 `dispatch_status` 明確 `CAST(:dispatch_status AS governance_dispatch_status)`，避免 enum/varchar 比較錯誤。
 - 補測 `test_ai_governance_endpoints.py`，覆蓋 dict remediation normalization 與 queue 查詢欄位相容性。
 
 **本地驗證**：

docs/superpowers/specs/2026-04-15-MASTER-ai-autonomous-flywheel-v2.md

@@ -2119,8 +2119,8 @@ Phase 6 完成後
 
 **T17b Governance API 紅燈修復（2026-05-14 台北）**：
 - 觸發：T17A 前端工作鏈路已能顯示治理卡點，但 live smoke 發現 `governance/events` 500、`governance/queue` 回 `table_pending=true`，會讓 Operator Console 無法可信呈現治理告警是否被 dispatch、跳過、修復或卡人工。
-- 根因：`ai_governance_events.details.remediation` 在 production 已有 dict/list 形態，read model 仍只收字串；`governance_remediation_dispatch` production schema 使用 `dispatched_at`，查詢卻讀不存在的 `created_at` / `operator_note`。
-- 修正：read-side normalization 將 remediation string/dict/list 正規化成短字串；queue query 改用 `d.dispatched_at AS created_at` 與 `NULL::text AS operator_note` 相容既有 DTO，不改 DB schema。
+- 根因：`ai_governance_events.details.remediation` 在 production 已有 dict/list 形態，read model 仍只收字串；`governance_remediation_dispatch` production schema 使用 `dispatched_at`，查詢卻讀不存在的 `created_at` / `operator_note`；第二層差異是 `dispatch_status` 為 PostgreSQL enum，未明確 cast 時會被 asyncpg 以 varchar 參數送入。
+- 修正：read-side normalization 將 remediation string/dict/list 正規化成短字串；queue query 改用 `d.dispatched_at AS created_at` 與 `NULL::text AS operator_note` 相容既有 DTO，並以 `CAST(:dispatch_status AS governance_dispatch_status)` 對齊 enum，不改 DB schema。
 - 驗證：`py_compile` pass；`pytest tests/test_ai_governance_endpoints.py tests/test_governance_remediation_dispatch.py -q` 53 passed；ruff F/E9 pass；diff check pass。
 - 目前進度更新：Alertmanager 低風險自動修復主線約 96%；完整 AI 自動化管理產品化約 75%。T17B 推版後，下一段收斂 governance dispatcher skipped reason / leader-dedupe / SLO emitter。