加入告警去重與洞察向量回補

services/event_router.py

@@ -22,6 +22,9 @@ _QUEUE_PATH = os.getenv(
     os.path.join(os.path.dirname(os.path.dirname(__file__)), "data", "event_router_failed_deliveries.jsonl"),
 )
 _QUEUE_LOCK = threading.Lock()
+_DEDUP_LOCK = threading.Lock()
+_EVENT_DEDUP: Dict[str, float] = {}
+_DEFAULT_DEDUP_SEC = int(os.getenv("MOMO_EVENT_ROUTER_DEFAULT_DEDUP_SEC", "0"))
 
 
 async def _handle_l1(event: Dict[str, Any], session_id: str) -> Dict[str, Any]:
@@ -112,6 +115,29 @@ def _is_event_silenced(event: Dict[str, Any]) -> bool:
         return False
 
 
+def _dedup_ttl_sec(event: Dict[str, Any]) -> int:
+    payload = event.get("payload") if isinstance(event.get("payload"), dict) else {}
+    raw = event.get("dedup_ttl_sec", payload.get("dedup_ttl_sec", _DEFAULT_DEDUP_SEC))
+    try:
+        return max(0, int(raw or 0))
+    except (TypeError, ValueError):
+        return 0
+
+
+def _is_duplicate_event(event: Dict[str, Any]) -> bool:
+    ttl = _dedup_ttl_sec(event)
+    if ttl <= 0:
+        return False
+    key = _event_key(event)
+    now = time.time()
+    with _DEDUP_LOCK:
+        until = _EVENT_DEDUP.get(key)
+        if until and until > now:
+            return True
+        _EVENT_DEDUP[key] = now + ttl
+    return False
+
+
 def _queue_failed_delivery(
     event: Dict[str, Any],
     tier: str,
@@ -197,6 +223,20 @@ async def dispatch(event: Dict[str, Any], admin_chat_ids: Optional[list] = None)
                 "delivered": True,
                 "silenced": True,
                 "queued": False,
+                "deduped": False,
+            }
+
+        if _is_duplicate_event(event):
+            return {
+                "tier": tier,
+                "sent": 0,
+                "errors": [],
+                "latency_ms": int((time.perf_counter() - started_at) * 1000),
+                "payload": {"status": "deduped", "event_key": _event_key(event)},
+                "delivered": True,
+                "silenced": False,
+                "queued": False,
+                "deduped": True,
             }
 
         result = await _run_tier_handler(tier, event, session_id)
@@ -220,6 +260,7 @@ async def dispatch(event: Dict[str, Any], admin_chat_ids: Optional[list] = None)
             "delivered": send_result["ok"],
             "silenced": False,
             "queued": queued,
+            "deduped": False,
         }
     except Exception as e:
         logger.exception(f"[EventRouter] dispatch failed: {e}")
@@ -233,6 +274,7 @@ async def dispatch(event: Dict[str, Any], admin_chat_ids: Optional[list] = None)
             "delivered": False,
             "silenced": False,
             "queued": queued,
+            "deduped": False,
         }
 
 
@@ -286,6 +328,7 @@ def notify_failure(
     title: Optional[str] = None,
     trace: Optional[str] = None,
     payload: Optional[Dict[str, Any]] = None,
+    dedup_ttl_sec: Optional[int] = None,
 ) -> Dict[str, Any]:
     """排程/背景任務失敗的同步通知 helper。"""
     severity = "alert" if priority in {"P1", "P2"} else "warning"
@@ -300,6 +343,8 @@ def notify_failure(
         "trace": trace or "".join(traceback.format_exception(type(error), error, error.__traceback__)),
         "payload": {"task_name": task_name, **(payload or {})},
     }
+    if dedup_ttl_sec is not None:
+        event["dedup_ttl_sec"] = dedup_ttl_sec
     return dispatch_sync(event)
 
 

services/openclaw_learning_service.py

@@ -62,6 +62,41 @@ def enqueue_insight_embedding(insight_id: int, insight_type: str, content: str,
     return _enqueue_embedding("ai_insights", int(insight_id), embed_target_text)
 
 
+def enqueue_missing_insight_embeddings(limit: int = 200) -> dict:
+    """Backfill existing ai_insights that have not yet entered the embedding queue."""
+    session = get_session()
+    try:
+        rows = session.execute(
+            text("""
+                SELECT id, insight_type, period, content
+                  FROM ai_insights i
+                 WHERE i.embedding IS NULL
+                   AND COALESCE(i.status, 'approved') NOT IN ('archived', 'rejected')
+                   AND NOT EXISTS (
+                       SELECT 1
+                         FROM embedding_retry_queue q
+                        WHERE q.target_table = 'ai_insights'
+                          AND q.target_id = i.id
+                          AND q.status IN ('pending', 'processing', 'done')
+                   )
+                 ORDER BY i.created_at DESC
+                 LIMIT :lim
+            """),
+            {"lim": limit},
+        ).fetchall()
+    except Exception as exc:
+        sys_log.warning(f"[OCLearn] embedding backfill 略過 (schema/pgvector 未就緒?): {exc}")
+        return {"scanned": 0, "enqueued": 0, "status": "skipped", "error": str(exc)[:200]}
+    finally:
+        session.close()
+
+    enqueued = 0
+    for row in rows:
+        if enqueue_insight_embedding(row.id, row.insight_type, row.content, row.period):
+            enqueued += 1
+    return {"scanned": len(rows), "enqueued": enqueued, "status": "ok"}
+
+
 def _process_one_embedding(row_id: int, target_table: str, target_id: int,
                            text_content: str, model: str) -> bool:
     """處理單筆 embedding，成功寫回目標表，失敗累加 attempts"""