fixed reasoning

backend/app/chat/service.py

@@ -37,6 +37,18 @@ MAX_HISTORY_MESSAGES = 40
 logger = logging.getLogger(__name__)
 
 
+def _build_messages_for_session(session_id: int) -> list[dict[str, Any]]:
+    db = SessionLocal()
+    try:
+        service = ChatService(db)
+        session = service.get_session(session_id)
+        if not session:
+            return []
+        return service._build_messages(session)
+    finally:
+        db.close()
+
+
 async def _extract_memory_background(
     session_id: int,
     user_text: str,
@@ -166,7 +178,12 @@ class ChatService:
             return
 
         self._save_message(session_id, "user", user_text)
-        messages = self._build_messages(session)
+        yield self._sse("status", {"phase": "preparing"})
+        messages = await asyncio.to_thread(_build_messages_for_session, session_id)
+        if not messages:
+            yield self._sse("error", {"message": "Session not found"})
+            return
+        yield self._sse("status", {"phase": "generating"})
         streamed_reply_parts: list[str] = []
 
         for _ in range(MAX_TOOL_ROUNDS):