fix Qdrant

backend/app/memory/service.py

@@ -1,7 +1,6 @@
 import asyncio
 import json
 import logging
-import threading
 from concurrent.futures import ThreadPoolExecutor
 from datetime import datetime, timezone
 from typing import Any
@@ -23,6 +22,8 @@ DEFAULT_PROFILE: dict[str, Any] = {
 
 logger = logging.getLogger(__name__)
 
+_RAG_EXECUTOR = ThreadPoolExecutor(max_workers=2, thread_name_prefix="rag")
+
 
 class MemoryService:
     def __init__(self, db: Session, user_id: int):
@@ -47,7 +48,7 @@ class MemoryService:
             except Exception:
                 logger.exception("RAG background task failed")
 
-        threading.Thread(target=runner, daemon=True).start()
+        _RAG_EXECUTOR.submit(runner)
 
     @staticmethod
     def _rag_fact_payload(fact: MemoryFact) -> dict[str, Any]:

backend/app/rag/store.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import logging
+from functools import lru_cache
 from typing import Any
 
 from qdrant_client import QdrantClient
@@ -14,11 +15,13 @@ COLLECTION_FACTS = "memory_facts"
 COLLECTION_SUMMARIES = "session_summaries"
 COLLECTION_DOC_CHUNKS = "document_chunks"
 VECTOR_SIZE = 1536
+QDRANT_TIMEOUT_SEC = 30.0
 
 
+@lru_cache
 def _client() -> QdrantClient:
     settings = get_settings()
-    return QdrantClient(url=settings.qdrant_url)
+    return QdrantClient(url=settings.qdrant_url, timeout=QDRANT_TIMEOUT_SEC)
 
 
 def ensure_collections() -> None: