fix(automation): run asset reconciliation on worker

apps/api/src/jobs/asset_capability_reconciliation_job.py

@@ -33,6 +33,8 @@ logger = structlog.get_logger(__name__)
 
 _FIRST_DELAY_SECONDS = 90
 _LOOP_BACKOFF_SECONDS = 30 * 60
+_MAX_RECONCILIATION_CANDIDATES = 20_000
+_RETRYABLE_RESULT_STATUSES = {"blocked_safe_no_write", "degraded_no_write"}
 _DAILY_TRIGGER_HOUR_TAIPEI = 3
 _DAILY_TRIGGER_MINUTE_TAIPEI = 30
 _TAIPEI = ZoneInfo("Asia/Taipei")
@@ -187,17 +189,30 @@ async def run_asset_capability_reconciliation_loop() -> None:
         daily_trigger_minute_taipei=_DAILY_TRIGGER_MINUTE_TAIPEI,
     )
     await asyncio.sleep(_FIRST_DELAY_SECONDS)
+    triggered_by = "startup"
     while True:
         try:
-            await reconcile_once()
+            result = await reconcile_once(triggered_by=triggered_by)
         except Exception as exc:
             logger.exception(
                 "asset_capability_reconciliation_loop_error",
                 error_type=type(exc).__name__,
             )
             await asyncio.sleep(_LOOP_BACKOFF_SECONDS)
+            triggered_by = "retry"
+            continue
+        if str(result.get("status") or "") in _RETRYABLE_RESULT_STATUSES:
+            logger.warning(
+                "asset_capability_reconciliation_retry_scheduled",
+                status=result.get("status"),
+                reason=result.get("reason"),
+                retry_seconds=_LOOP_BACKOFF_SECONDS,
+            )
+            await asyncio.sleep(_LOOP_BACKOFF_SECONDS)
+            triggered_by = "retry"
             continue
         await asyncio.sleep(_seconds_until_next_trigger())
+        triggered_by = "cron"
 
 
 async def reconcile_once(
@@ -231,6 +246,19 @@ async def reconcile_once(
         }
 
     candidates = build_reconciliation_candidates(matrix)
+    if len(candidates) > _MAX_RECONCILIATION_CANDIDATES:
+        logger.error(
+            "asset_capability_reconciliation_candidate_bound_exceeded",
+            candidate_count=len(candidates),
+            max_candidate_count=_MAX_RECONCILIATION_CANDIDATES,
+            external_runtime_write_performed=False,
+        )
+        return {
+            "status": "blocked_safe_no_write",
+            "reason": "candidate_count_above_controlled_execution_bound",
+            "candidate_count": len(candidates),
+            "max_candidate_count": _MAX_RECONCILIATION_CANDIDATES,
+        }
     receipt = await _persist_reconciliation(
         matrix,
         candidates,
@@ -284,6 +312,8 @@ async def _persist_reconciliation(
                   AND COALESCE(input ->> 'semantic_operation_type', operation_type)
                       = :semantic_operation_type
                   AND input ->> 'idempotency_key' = :idempotency_key
+                  AND status = 'success'
+                  AND output ->> 'closed' = 'true'
                 ORDER BY created_at DESC
                 LIMIT 1
                 """
@@ -298,28 +328,38 @@ async def _persist_reconciliation(
         if isinstance(prior, dict):
             return prior
 
-        for candidate_raw in candidates:
-            candidate = dict(candidate_raw)
+        candidate_fingerprints = [
+            str(candidate.get("candidate_fingerprint") or "")
+            for candidate in candidates
+        ]
+        existing_fingerprints: set[str] = set()
+        if candidate_fingerprints:
             existing = await db.execute(
                 text(
                     """
-                    SELECT op_id::text
+                    SELECT DISTINCT input ->> 'candidate_fingerprint'
                     FROM automation_operation_log
                     WHERE actor = :actor
                       AND COALESCE(input ->> 'semantic_operation_type', operation_type)
                           = :semantic_operation_type
-                      AND input ->> 'candidate_fingerprint' = :candidate_fingerprint
-                    ORDER BY created_at DESC
-                    LIMIT 1
+                      AND input ->> 'candidate_fingerprint'
+                          = ANY(CAST(:candidate_fingerprints AS text[]))
                     """
                 ),
                 {
                     "actor": RECONCILIATION_ACTOR,
                     "semantic_operation_type": RECONCILIATION_WORK_ITEM_OPERATION,
-                    "candidate_fingerprint": candidate["candidate_fingerprint"],
+                    "candidate_fingerprints": candidate_fingerprints,
                 },
             )
-            if existing.scalar_one_or_none():
+            existing_fingerprints = {
+                str(row[0]) for row in existing.fetchall() if row[0]
+            }
+
+        insert_rows: list[dict[str, Any]] = []
+        for candidate_raw in candidates:
+            candidate = dict(candidate_raw)
+            if candidate["candidate_fingerprint"] in existing_fingerprints:
                 continue
             input_payload = build_reconciliation_work_item_input(
                 candidate,
@@ -334,6 +374,29 @@ async def _persist_reconciliation(
                 "post_verifier": "pending_batch_readback",
                 "rollback_or_no_write_terminal": "no_external_runtime_write",
             }
+            insert_rows.append(
+                {
+                    "asset_id": candidate.get("asset_inventory_id"),
+                    "discovery_run_id": discovery_run_id,
+                    "actor": RECONCILIATION_ACTOR,
+                    "input": _canonical_json(input_payload),
+                    "output": _canonical_json(output_payload),
+                    "dry_run_result": _canonical_json(
+                        {
+                            "would_create_work_item": True,
+                            "external_runtime_write": False,
+                            "candidate_fingerprint": candidate["candidate_fingerprint"],
+                        }
+                    ),
+                    "tags": [
+                        "asset_capability_reconciliation",
+                        "work_item",
+                        "aia_p0_006",
+                    ],
+                }
+            )
+
+        if insert_rows:
             await db.execute(
                 text(
                     """
@@ -360,50 +423,37 @@ async def _persist_reconciliation(
                     )
                     """
                 ),
-                {
-                    "asset_id": candidate.get("asset_inventory_id"),
-                    "discovery_run_id": discovery_run_id,
-                    "actor": RECONCILIATION_ACTOR,
-                    "input": _canonical_json(input_payload),
-                    "output": _canonical_json(output_payload),
-                    "dry_run_result": _canonical_json(
-                        {
-                            "would_create_work_item": True,
-                            "external_runtime_write": False,
-                            "candidate_fingerprint": candidate["candidate_fingerprint"],
-                        }
-                    ),
-                    "tags": [
-                        "asset_capability_reconciliation",
-                        "work_item",
-                        "aia_p0_006",
-                    ],
-                },
+                insert_rows,
             )
-            created_count += 1
+            created_count = len(insert_rows)
 
-        rows_result = await db.execute(
-            text(
-                """
-                SELECT input ->> 'candidate_fingerprint' AS candidate_fingerprint
-                FROM automation_operation_log
-                WHERE actor = :actor
-                  AND COALESCE(input ->> 'semantic_operation_type', operation_type)
-                      = :semantic_operation_type
-                """
-            ),
-            {
-                "actor": RECONCILIATION_ACTOR,
-                "semantic_operation_type": RECONCILIATION_WORK_ITEM_OPERATION,
-            },
-        )
-        persisted_fingerprints = {
-            str(row[0]) for row in rows_result.fetchall() if row[0]
-        }
         expected_fingerprints = {
             str(candidate.get("candidate_fingerprint") or "")
             for candidate in candidates
         }
+        persisted_fingerprints: set[str] = set()
+        if candidate_fingerprints:
+            rows_result = await db.execute(
+                text(
+                    """
+                    SELECT DISTINCT input ->> 'candidate_fingerprint'
+                    FROM automation_operation_log
+                    WHERE actor = :actor
+                      AND COALESCE(input ->> 'semantic_operation_type', operation_type)
+                          = :semantic_operation_type
+                      AND input ->> 'candidate_fingerprint'
+                          = ANY(CAST(:candidate_fingerprints AS text[]))
+                    """
+                ),
+                {
+                    "actor": RECONCILIATION_ACTOR,
+                    "semantic_operation_type": RECONCILIATION_WORK_ITEM_OPERATION,
+                    "candidate_fingerprints": candidate_fingerprints,
+                },
+            )
+            persisted_fingerprints = {
+                str(row[0]) for row in rows_result.fetchall() if row[0]
+            }
         repository_readback_count = len(expected_fingerprints & persisted_fingerprints)
         summary_output = build_reconciliation_summary_payload(
             matrix,

apps/api/src/main.py

@@ -191,6 +191,7 @@ def _resolve_request_project_context(request: Request) -> tuple[str | None, str]
 
     return None, "request.project_id.missing"
 
+
 # =============================================================================
 # Sentry SDK Initialization (Error Tracking - 補強 SignOz)
 # Self-Hosted @ 192.168.0.110
@@ -274,6 +275,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # AwoooP Phase 2.4 (2026-05-04 ogt): 設定 startup handler 的 project_id context
     # asyncio.create_task() 自動繼承父任務的 ContextVar → 31 個 background loop 全部標記為 awoooi
     from src.core.context import PROJECT_ID
+
     PROJECT_ID.set("awoooi")
 
     # Startup
@@ -345,6 +347,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
 
     # ADR-015: MCP Provider 註冊 (DI 模式)
     from src.plugins.mcp.providers import register_all_providers
+
     register_all_providers()
     logger.info("mcp_providers_registered")
 
@@ -352,6 +355,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-16 ogt + Claude Sonnet 4.6: 修復 sensors=0/0 根因 — init 從未在 startup 被呼叫
     try:
         from src.services.mcp_tool_registry import init_mcp_tool_registry
+
         await init_mcp_tool_registry()
         logger.info("mcp_tool_registry_initialized")
     except Exception as e:
@@ -367,7 +371,9 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         )
         logger.info("telegram_heartbeat_monitor_started", interval_minutes=30)
     else:
-        logger.warning("telegram_heartbeat_monitor_skipped", reason="OPENCLAW_TG_BOT_TOKEN not set")
+        logger.warning(
+            "telegram_heartbeat_monitor_skipped", reason="OPENCLAW_TG_BOT_TOKEN not set"
+        )
 
     # Reboot Recovery: Warm-up Redis Working Memory from PostgreSQL
     # 2026-04-05 ogt: 重開機後 Redis 清空，從 DB restore 未解決的 incidents
@@ -392,10 +398,12 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
             async with get_db_context() as db:
                 result = await db.execute(
                     select(IncidentRecord).where(
-                        IncidentRecord.status.in_([
-                            IncidentStatus.INVESTIGATING,
-                            IncidentStatus.MITIGATING,
-                        ])
+                        IncidentRecord.status.in_(
+                            [
+                                IncidentStatus.INVESTIGATING,
+                                IncidentStatus.MITIGATING,
+                            ]
+                        )
                     )
                 )
                 records = result.scalars().all()
@@ -446,6 +454,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 必須在 embedding indexing 之前，確保 playbook 表有資料
     try:
         from src.services.playbook_seed_service import seed_playbooks_from_rules
+
         schedule_api_background_task(seed_playbooks_from_rules())
         logger.info("playbook_seed_scheduled")
     except Exception as e:
@@ -456,6 +465,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 3.5 AI 學習成果持久化
     try:
         from src.repositories.playbook_repository import get_playbook_repository
+
         schedule_api_background_task(get_playbook_repository().backfill_redis_to_pg())
         logger.info("playbook_pg_backfill_scheduled")
     except Exception as e:
@@ -465,6 +475,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         from src.services.playbook_embedding_service import (
             ensure_playbook_embeddings_indexed,
         )
+
         schedule_api_background_task(ensure_playbook_embeddings_indexed())
         logger.info("playbook_embedding_indexing_scheduled")
     except Exception as e:
@@ -486,6 +497,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # dedup TTL 10 分鐘過期後，ready decisions 就沒有補送機制 → 長期卡在 ready 無人審核
     try:
         from src.services.decision_manager import get_decision_manager
+
         schedule_api_background_task(get_decision_manager().resend_stale_ready_tokens())
         logger.info("stale_ready_tokens_resend_scheduled")
     except Exception as e:
@@ -497,6 +509,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # Sweeper 定期掃描無 decision token 的 INVESTIGATING incidents → 背景觸發
     try:
         from src.jobs.incident_analysis_sweeper import run_incident_analysis_sweeper
+
         schedule_api_background_task(run_incident_analysis_sweeper())
         logger.info("incident_analysis_sweeper_scheduled", interval_sec=90)
     except Exception as e:
@@ -507,6 +520,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 解開 8 張 0 writer 表的第一個 (asset_inventory / asset_discovery_run / asset_coverage_snapshot)
     try:
         from src.jobs.asset_scanner_job import run_asset_scanner_loop
+
         schedule_api_background_task(run_asset_scanner_loop())
         logger.info("asset_scanner_loop_scheduled", interval_sec=3600)
     except Exception as e:
@@ -517,6 +531,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 解鎖 E3 Hermes 自動建規則: AI 需要 alert_rule_catalog 作為 baseline 才能提案修正
     try:
         from src.jobs.rule_catalog_sync_job import run_rule_catalog_sync_loop
+
         schedule_api_background_task(run_rule_catalog_sync_loop())
         logger.info("rule_catalog_sync_loop_scheduled", interval_sec=3600)
     except Exception as e:
@@ -527,6 +542,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 解鎖: Phase 4 Holt-Winters 預測有歷史資料 / 容量趨勢分析
     try:
         from src.jobs.capacity_scanner_job import run_capacity_scanner_loop
+
         schedule_api_background_task(run_capacity_scanner_loop())
         logger.info("capacity_scanner_loop_scheduled", daily_trigger_hour_taipei=2)
     except Exception as e:
@@ -537,6 +553,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # MVP: secret_rotated 真實檢查,其他 6 維占位 'unknown',後續 agent 補
     try:
         from src.jobs.compliance_scanner_job import run_compliance_scanner_loop
+
         schedule_api_background_task(run_compliance_scanner_loop())
         logger.info("compliance_scanner_loop_scheduled", daily_trigger_hour_taipei=3)
     except Exception as e:
@@ -546,6 +563,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 消費 signals:aider:events stream → 建 incident + 寫 aider_events 表
     try:
         from src.workers.aider_event_processor import run_aider_event_processor_loop
+
         logger.info("aider_event_processor: starting Redis stream consumer")
         schedule_api_background_task(run_aider_event_processor_loop())
     except Exception as e:
@@ -556,6 +574,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 依據: Prometheus targets / alert_rule_catalog labels / knowledge_entries 覆蓋
     try:
         from src.jobs.coverage_evaluator_job import run_coverage_evaluator_loop
+
         schedule_api_background_task(run_coverage_evaluator_loop())
         logger.info("coverage_evaluator_loop_scheduled", interval_sec=3600)
     except Exception as e:
@@ -566,6 +585,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 解鎖 E3 Hermes: noise_rate > 0.5 的 rule 可被 AI 提案 deprecate
     try:
         from src.jobs.rule_stats_updater_job import run_rule_stats_updater_loop
+
         schedule_api_background_task(run_rule_stats_updater_loop())
         logger.info("rule_stats_updater_loop_scheduled", interval_sec=3600)
     except Exception as e:
@@ -576,35 +596,26 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 解鎖: 資產變化歷史 (added/removed/lifecycle_changed),AI 可追蹤集群演進
     try:
         from src.jobs.asset_change_tracker_job import run_asset_change_tracker_loop
+
         schedule_api_background_task(run_asset_change_tracker_loop())
         logger.info("asset_change_tracker_loop_scheduled", interval_sec=3600)
     except Exception as e:
         logger.warning("asset_change_tracker_loop_schedule_failed", error=str(e))
 
-    # AIA-P0-006: reconcile the declared all-asset catalog with live ADR-090
-    # inventory and project added/removed/drift capability gaps as durable work items.
-    try:
-        from src.jobs.asset_capability_reconciliation_job import (
-            run_asset_capability_reconciliation_loop,
-        )
-
-        schedule_api_background_task(run_asset_capability_reconciliation_loop())
-        logger.info(
-            "asset_capability_reconciliation_loop_scheduled",
-            daily_trigger_hour_taipei=3,
-            daily_trigger_minute_taipei=30,
-        )
-    except Exception as e:
-        logger.warning(
-            "asset_capability_reconciliation_loop_schedule_failed",
-            error=str(e),
-        )
+    # AIA-P0-006 has one runtime owner. Production API pods reserve their DB
+    # budget for requests, so the singleton signal worker owns reconciliation.
+    logger.info(
+        "asset_capability_reconciliation_owner_delegated",
+        owner="signal_worker",
+        reason="api_db_pool_reserved_for_serving_requests",
+    )
 
     # ADR-090 § Hermes Rule Quality Advisor (2026-04-19 ogt + Claude Opus 4.7 Asia/Taipei)
     # 每日 04:00 Taipei 分析 alert_rule_catalog.noise_rate,對高噪音規則推 Telegram 建議
     # 統帥鐵律: AI 只推建議不自動改 review_status,人工決策 deprecate
     try:
         from src.jobs.hermes_rule_quality_job import run_hermes_rule_quality_loop
+
         schedule_api_background_task(run_hermes_rule_quality_loop())
         logger.info("hermes_rule_quality_loop_scheduled", daily_trigger_hour_taipei=4)
     except Exception as e:
@@ -615,6 +626,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 高風險 host 寫 aol(capacity_recommendation) + Telegram 建議
     try:
         from src.jobs.capacity_forecaster_job import run_capacity_forecaster_loop
+
         schedule_api_background_task(run_capacity_forecaster_loop())
         logger.info("capacity_forecaster_loop_scheduled", daily_trigger_hour_taipei=5)
     except Exception as e:
@@ -628,6 +640,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
             run_monthly_report_loop,
             run_weekly_report_loop,
         )
+
         schedule_api_background_task(run_daily_report_loop())
         schedule_api_background_task(run_weekly_report_loop())
         schedule_api_background_task(run_monthly_report_loop())
@@ -644,6 +657,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 確保 PENDING approval 超過 48h 後觸發 resolve_incident → KM 學習鏈閉環
     try:
         from src.jobs.approval_timeout_resolver import run_approval_timeout_resolver
+
         schedule_api_background_task(run_approval_timeout_resolver())
         logger.info("approval_timeout_resolver_scheduled", interval_sec=3600)
     except Exception as e:
@@ -657,6 +671,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         from src.jobs.incident_lifecycle_reconciler import (
             INTERVAL_SECONDS as INCIDENT_LIFECYCLE_RECONCILER_INTERVAL,
         )
+
         logger.info(
             "incident_lifecycle_reconciler_owner_delegated",
             interval_sec=INCIDENT_LIFECYCLE_RECONCILER_INTERVAL,
@@ -676,6 +691,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         from src.jobs.awooop_ansible_candidate_backfill_job import (
             run_awooop_ansible_candidate_backfill_loop,
         )
+
         scheduled_task = schedule_api_background_task(
             run_awooop_ansible_candidate_backfill_loop()
         )
@@ -687,7 +703,9 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
             interval_seconds=settings.AWOOOP_ANSIBLE_CANDIDATE_BACKFILL_INTERVAL_SECONDS,
         )
     except Exception as e:
-        logger.warning("awooop_ansible_candidate_backfill_worker_schedule_failed", error=str(e))
+        logger.warning(
+            "awooop_ansible_candidate_backfill_worker_schedule_failed", error=str(e)
+        )
 
     # AwoooP Ansible check-mode worker.
     # 先執行 ansible-playbook --check --diff 並回寫 automation_operation_log；
@@ -696,6 +714,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         from src.jobs.awooop_ansible_check_mode_job import (
             run_awooop_ansible_check_mode_loop,
         )
+
         scheduled_task = schedule_api_background_task(
             run_awooop_ansible_check_mode_loop()
         )
@@ -713,6 +732,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 3 初始建立
     try:
         from src.services.playbook_evolver import run_evolver_loop
+
         schedule_api_background_task(run_evolver_loop())
         logger.info("evolver_loop_scheduled", interval_sec=86400)
     except Exception as e:
@@ -723,18 +743,22 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         from src.jobs.playbook_generation_governance_job import (
             run_playbook_generation_governance_loop,
         )
+
         schedule_api_background_task(run_playbook_generation_governance_loop())
         logger.info(
             "playbook_generation_governance_loop_scheduled",
             interval_sec=settings.PLAYBOOK_DRAFT_GOVERNANCE_INTERVAL_SECONDS,
         )
     except Exception as e:
-        logger.warning("playbook_generation_governance_loop_schedule_failed", error=str(e))
+        logger.warning(
+            "playbook_generation_governance_loop_schedule_failed", error=str(e)
+        )
 
     # ADR-083 Phase 3: 知識遺忘 Job（每日）— 30d 未引用 KB entry 標記 archived
     # 2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 3 初始建立
     try:
         from src.jobs.knowledge_decay_job import run_knowledge_decay_loop
+
         schedule_api_background_task(run_knowledge_decay_loop())
         logger.info("knowledge_decay_loop_scheduled", interval_sec=86400)
     except Exception as e:
@@ -745,6 +769,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # Feature Flag: ENABLE_KM_BACKFILL_RECONCILER=false 停用（回滾用）
     try:
         from src.jobs.km_backfill_reconciler_job import run_km_backfill_reconciler_loop
+
         schedule_api_background_task(run_km_backfill_reconciler_loop())
         logger.info("km_backfill_reconciler_loop_scheduled", interval_sec=300)
     except Exception as e:
@@ -756,6 +781,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # ADR-091 Task T2
     try:
         from src.jobs.aol_to_catalog_writeback_job import run_aol_writeback_loop
+
         schedule_api_background_task(run_aol_writeback_loop())
         logger.info("aol_to_catalog_writeback_loop_scheduled", interval_sec=3600)
     except Exception as e:
@@ -771,28 +797,48 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
             import asyncio as _asyncio
 
             from src.jobs.kb_rot_cleaner import get_kb_rot_cleaner
+
             while True:
                 try:
                     now = now_taipei()
                     # 計算下次月初 3 點（台北時間）
                     if now.day == 1 and now.hour < 3:
-                        next_run = now.replace(hour=3, minute=0, second=0, microsecond=0)
+                        next_run = now.replace(
+                            hour=3, minute=0, second=0, microsecond=0
+                        )
                     elif now.month == 12:
                         next_run = now.replace(
-                            year=now.year + 1, month=1, day=1,
-                            hour=3, minute=0, second=0, microsecond=0,
+                            year=now.year + 1,
+                            month=1,
+                            day=1,
+                            hour=3,
+                            minute=0,
+                            second=0,
+                            microsecond=0,
                         )
                     else:
                         next_run = now.replace(
-                            month=now.month + 1, day=1,
-                            hour=3, minute=0, second=0, microsecond=0,
+                            month=now.month + 1,
+                            day=1,
+                            hour=3,
+                            minute=0,
+                            second=0,
+                            microsecond=0,
                         )
                     sleep_sec = (next_run - now).total_seconds()
-                    logger.info("kb_rot_cleaner_next_run", next_run=next_run.isoformat(), sleep_sec=int(sleep_sec))
+                    logger.info(
+                        "kb_rot_cleaner_next_run",
+                        next_run=next_run.isoformat(),
+                        sleep_sec=int(sleep_sec),
+                    )
                     await _asyncio.sleep(sleep_sec)
                     try:
                         result = await get_kb_rot_cleaner().run()
-                        logger.info("kb_rot_cleaner_completed", stale_count=result.stale_count, total=result.total_scanned)
+                        logger.info(
+                            "kb_rot_cleaner_completed",
+                            stale_count=result.stale_count,
+                            total=result.total_scanned,
+                        )
                     except Exception as _e:
                         logger.exception("kb_rot_cleaner_failed", error=str(_e))
                 except _asyncio.CancelledError:
@@ -810,6 +856,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 3 初始建立
     try:
         from src.services.finetune_exporter import run_finetune_export_loop
+
         schedule_api_background_task(run_finetune_export_loop())
         logger.info("finetune_export_loop_scheduled", interval_sec=604800)
     except Exception as e:
@@ -821,6 +868,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 4 初始建立
     try:
         from src.services.proactive_inspector import run_proactive_inspector_loop
+
         schedule_api_background_task(run_proactive_inspector_loop())
         logger.info("proactive_inspector_loop_scheduled", interval_sec=300)
     except Exception as e:
@@ -830,6 +878,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-04-15 ogt + Claude Sonnet 4.6（亞太）: Phase 6 初始建立
     try:
         from src.jobs.offline_replay_service import run_offline_replay_loop
+
         schedule_api_background_task(run_offline_replay_loop())
         logger.info("offline_replay_loop_scheduled", interval_sec=604800)
     except Exception as e:
@@ -839,6 +888,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # MASTER §1.1：系統必須能感知自身故障，W-1 SLO + W-2 Telegram靜默 + W-3 飛輪成功率
     try:
         from src.jobs.ai_slo_watchdog_job import run_ai_slo_watchdog_loop
+
         schedule_api_background_task(run_ai_slo_watchdog_loop())
         logger.info("ai_slo_watchdog_scheduled", interval_sec=900)
     except Exception as e:
@@ -848,6 +898,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # MASTER P2.2：trust_drift / knowledge_degradation / llm_hallucination / execution_blast_radius
     try:
         from src.services.governance_agent import run_governance_loop
+
         schedule_api_background_task(run_governance_loop())
         logger.info("governance_agent_scheduled", interval_sec=3600)
     except Exception as e:
@@ -856,6 +907,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 2026-05-03 ogt + Claude Sonnet 4.6（亞太）: GovernanceDispatcher Wave 2E（每 30s poll）
     try:
         from src.services.governance_dispatcher import run_governance_dispatcher_loop
+
         schedule_api_background_task(run_governance_dispatcher_loop())
         logger.info("governance_dispatcher_scheduled", interval_sec=30)
     except Exception as e:
@@ -866,6 +918,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 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
+
         schedule_api_background_task(run_hermes_kb_growth_loop())
         logger.info("hermes_kb_growth_worker_scheduled", interval_sec=300)
     except Exception as e:
@@ -893,6 +946,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
         try:
             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:
@@ -909,8 +963,10 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 探測 5 Provider（ollama/ollama_local/gemini/claude/openclaw_nemo）版本
     # 寫入 ai_provider_version_history；版本變更時 log warning，P3.2.3 alerter 後續整合
     try:
+
         async def _run_model_version_tracker_loop() -> None:
             from src.services.model_version_tracker import get_model_version_tracker
+
             tracker = get_model_version_tracker()
             while True:
                 try:
@@ -924,7 +980,9 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
                 except asyncio.CancelledError:
                     break
                 except Exception as _loop_err:
-                    logger.exception("model_version_tracker_loop_error", error=str(_loop_err))
+                    logger.exception(
+                        "model_version_tracker_loop_error", error=str(_loop_err)
+                    )
                     await asyncio.sleep(60)  # 錯誤後 1 分鐘重試
 
         schedule_api_background_task(_run_model_version_tracker_loop())
@@ -937,6 +995,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # Shadow mode：is_shadow=True，0 user-visible response，0 destructive tool call
     try:
         from src.workers.platform_worker import start_platform_worker
+
         if api_background_tasks_enabled:
             await start_platform_worker()
             logger.info("platform_worker_started", mode="shadow")
@@ -955,6 +1014,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 必須在 Redis pool 關閉之前停止（recovery service 可能仍在寫 Redis）
     try:
         from src.services.ollama_auto_recovery import get_ollama_auto_recovery_service
+
         await get_ollama_auto_recovery_service().stop()
         logger.info("ollama_failover_system_stopped")
     except Exception as e:
@@ -965,6 +1025,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # 確保 _verify_and_learn / runbook_generator 寫入不被 SIGTERM cancel
     try:
         from src.services.auto_repair_service import get_auto_repair_service
+
         _repair_svc = get_auto_repair_service()
         if hasattr(_repair_svc, "drain_pending_tasks"):
             _drain_result = await _repair_svc.drain_pending_tasks(timeout=60.0)
@@ -975,6 +1036,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     # AwoooP Phase 4: Platform Worker 優雅停機（2026-05-04 ogt）
     try:
         from src.workers.platform_worker import stop_platform_worker
+
         await stop_platform_worker()
         logger.info("platform_worker_stopped")
     except Exception as e:
@@ -991,6 +1053,7 @@ async def lifespan(_app: FastAPI) -> AsyncGenerator[None, None]:
     await telegram_gw.close()
     # Phase 33: Close RAG Service httpx client (ADR-067)
     from src.services.knowledge_rag_service import get_knowledge_rag_service
+
     await get_knowledge_rag_service().close()
     # Phase 5: Close HTTP Clients (統帥鐵律: 連線池回收)
     await close_all_http_clients()
@@ -1178,7 +1241,9 @@ async def http_exception_handler(_request: Request, exc: HTTPException) -> JSONR
     This is critical for P1-1 fail-closed evidence; without it, all HTTPException
     is swallowed by the generic exception handler and downgraded to 500.
     """
-    return JSONResponse(status_code=exc.status_code, content={"detail": exc.detail}, headers=exc.headers)
+    return JSONResponse(
+        status_code=exc.status_code, content={"detail": exc.detail}, headers=exc.headers
+    )
 
 
 @app.exception_handler(Exception)
@@ -1219,15 +1284,25 @@ app.include_router(csrf_v1.router, prefix="/api/v1", tags=["Security"])  # Phase
 app.include_router(dashboard_v1.router, prefix="/api/v1", tags=["Dashboard"])
 app.include_router(approvals_v1.router, prefix="/api/v1", tags=["HITL Approvals"])
 app.include_router(ai_v1.router, prefix="/api/v1", tags=["AI Decision"])
-app.include_router(ai_governance_v1.router, prefix="/api/v1", tags=["AI Governance"])  # 2026-05-02: /governance 頁面
-app.include_router(ai_slo_v1.router, prefix="/api/v1", tags=["AI SLO"])  # Phase 6 ADR-087
-app.include_router(aiops_kpi_v1.router, prefix="/api/v1", tags=["AIOps KPI"])  # ADR-090 § Phase 7 Dashboard
-app.include_router(aiops_timeline_v1.router, prefix="/api/v1", tags=["AIOps Timeline"])  # 2026-04-27 Wave8-X3 B4
+app.include_router(
+    ai_governance_v1.router, prefix="/api/v1", tags=["AI Governance"]
+)  # 2026-05-02: /governance 頁面
+app.include_router(
+    ai_slo_v1.router, prefix="/api/v1", tags=["AI SLO"]
+)  # Phase 6 ADR-087
+app.include_router(
+    aiops_kpi_v1.router, prefix="/api/v1", tags=["AIOps KPI"]
+)  # ADR-090 § Phase 7 Dashboard
+app.include_router(
+    aiops_timeline_v1.router, prefix="/api/v1", tags=["AIOps Timeline"]
+)  # 2026-04-27 Wave8-X3 B4
 app.include_router(webhooks_v1.router, prefix="/api/v1", tags=["Webhooks"])
 app.include_router(timeline_v1.router, prefix="/api/v1", tags=["Timeline"])
 app.include_router(audit_logs_v1.router, prefix="/api/v1", tags=["Audit Logs"])
 # 2026-04-09 Claude Sonnet 4.6: alert_operation_log 查詢 API (Sprint 5.2)
-app.include_router(alert_operation_logs_v1.router, prefix="/api/v1", tags=["Alert Operation Logs"])
+app.include_router(
+    alert_operation_logs_v1.router, prefix="/api/v1", tags=["Alert Operation Logs"]
+)
 app.include_router(
     aider_events_v1.router,
     prefix="/api/v1",
@@ -1310,7 +1385,9 @@ app.include_router(
     notifications.router, prefix="/api/v1/notifications", tags=["Notifications"]
 )
 # AwoooP Phase 4 (2026-05-04 ogt): Platform Shell — Shadow Mode Run API
-app.include_router(platform_v1.router, prefix="/api/v1/platform", tags=["AwoooP Platform"])
+app.include_router(
+    platform_v1.router, prefix="/api/v1/platform", tags=["AwoooP Platform"]
+)
 
 
 # =============================================================================

apps/api/src/services/ai_automation_asset_capability_matrix.py

@@ -631,6 +631,7 @@ def _receipt_controls(
     fresh = bool(
         row
         and str(row.get("status") or "") == "success"
+        and output.get("closed") is True
         and _is_fresh(
             created_at,
             now=now,

apps/api/src/workers/signal_worker.py

@@ -71,6 +71,7 @@ GRACEFUL_SHUTDOWN_TIMEOUT_S = 75  # K8s terminationGracePeriodSeconds: 90
 # Signal Worker
 # =============================================================================
 
+
 class SignalWorker:
     """
     Redis Streams 訊號消費者
@@ -366,7 +367,9 @@ class SignalWorker:
                 span.set_attribute("messaging.destination", STREAM_KEY)
                 span.set_attribute("messaging.message_id", message_id)
                 span.set_attribute("signal.source", data.get("source", "unknown"))
-                span.set_attribute("signal.alert_name", data.get("alert_name", "unknown"))
+                span.set_attribute(
+                    "signal.alert_name", data.get("alert_name", "unknown")
+                )
                 span.set_attribute("signal.severity", data.get("severity", "unknown"))
 
                 logger.info(
@@ -400,7 +403,9 @@ class SignalWorker:
                         severity=incident.severity.value,
                         signal_count=len(incident.signals),
                         affected_services=incident.affected_services,
-                        persisted_to_pg=getattr(incident, "persisted_to_pg", False),  # 2026-04-01 ogt: BrainIncident 無此欄位 (ADR-046 P2-01)
+                        persisted_to_pg=getattr(
+                            incident, "persisted_to_pg", False
+                        ),  # 2026-04-01 ogt: BrainIncident 無此欄位 (ADR-046 P2-01)
                     )
                     try:
                         from src.services.signal_observation_service import (
@@ -505,6 +510,7 @@ def get_signal_worker() -> SignalWorker:
 # Standalone Entry Point (for K8s Worker Deployment)
 # =============================================================================
 
+
 async def _write_health_files() -> None:
     """
     Write health check files for K8s probes.
@@ -542,10 +548,7 @@ async def _heartbeat_loop(shutdown_event: asyncio.Event) -> None:
 
         # 等待下次心跳或收到關閉信號
         try:
-            await asyncio.wait_for(
-                shutdown_event.wait(),
-                timeout=HEARTBEAT_INTERVAL
-            )
+            await asyncio.wait_for(shutdown_event.wait(), timeout=HEARTBEAT_INTERVAL)
             break  # 收到關閉信號
         except TimeoutError:
             continue  # 超時，繼續下次心跳
@@ -575,7 +578,9 @@ async def _main() -> None:
         "signal_worker_standalone_starting",
         environment=settings.ENVIRONMENT,
         redis_url=settings.REDIS_URL.split("@")[-1] if settings.REDIS_URL else "N/A",
-        database_url=settings.DATABASE_URL.split("@")[-1] if settings.DATABASE_URL else "N/A",
+        database_url=settings.DATABASE_URL.split("@")[-1]
+        if settings.DATABASE_URL
+        else "N/A",
     )
 
     # Initialize Redis (API pool + Worker 專屬長連線池)
@@ -695,6 +700,9 @@ async def _main() -> None:
                 telegram_dispatch_performed=False,
             )
     if settings.ENABLE_SECURITY_CONTROL_PLANE_MAINTENANCE_WORKER:
+        from src.jobs.asset_capability_reconciliation_job import (
+            run_asset_capability_reconciliation_loop,
+        )
         from src.jobs.asset_change_tracker_job import run_asset_change_tracker_loop
         from src.jobs.asset_scanner_job import run_asset_scanner_loop
         from src.jobs.compliance_scanner_job import run_compliance_scanner_loop
@@ -707,6 +715,10 @@ async def _main() -> None:
             ("run_compliance_scanner_loop", run_compliance_scanner_loop),
             ("run_coverage_evaluator_loop", run_coverage_evaluator_loop),
             ("run_asset_change_tracker_loop", run_asset_change_tracker_loop),
+            (
+                "run_asset_capability_reconciliation_loop",
+                run_asset_capability_reconciliation_loop,
+            ),
         )
         security_maintenance_tasks = [
             asyncio.create_task(loop(), name=task_name)
@@ -791,12 +803,8 @@ async def _main() -> None:
             "signal_worker_agent99_controlled_dispatch_reconciler_skipped_fail_closed",
             owner="signal_worker",
             reason="authenticated_relay_not_configured",
-            relay_url_configured=bool(
-                settings.AGENT99_SRE_ALERT_RELAY_URL.strip()
-            ),
-            relay_token_configured=bool(
-                settings.AGENT99_SRE_ALERT_RELAY_TOKEN.strip()
-            ),
+            relay_url_configured=bool(settings.AGENT99_SRE_ALERT_RELAY_URL.strip()),
+            relay_token_configured=bool(settings.AGENT99_SRE_ALERT_RELAY_TOKEN.strip()),
         )
     # Setup graceful shutdown
     shutdown_event = asyncio.Event()
@@ -873,6 +881,7 @@ async def _main() -> None:
 
     # Remove health files
     from pathlib import Path
+
     Path("/tmp/worker-healthy").unlink(missing_ok=True)
     Path("/tmp/worker-ready").unlink(missing_ok=True)