修改rag，实现混合检索

backend/app/backend.py

@@ -98,7 +98,7 @@ async def health_check():
 class ChatRequest(BaseModel):
     message: str
     thread_id: str | None = None
-    model: str = "zhipu"
+    model: str = "local"
     user_id: str = "default_user"
 
 class ChatResponse(BaseModel):
@@ -212,7 +212,7 @@ async def chat_endpoint(
 @app.get("/threads")
 async def list_threads(
     user_id: str = Query("default_user", description="用户 ID"),
-    limit: int = Query(50, ge=1, le=200, description="返回数量限制"),
+    limit: int = Query(4, ge=1, le=200, description="返回数量限制"),
     history_service: ThreadHistoryService = Depends(get_history_service)
 ):
     """获取当前用户的对话历史列表"""
@@ -312,7 +312,7 @@ async def websocket_endpoint(
             data = await websocket.receive_json()
             message = data.get("message")
             thread_id = data.get("thread_id", str(uuid.uuid4()))
-            model = data.get("model", "zhipu")
+            model = data.get("model", "local")
             user_id = data.get("user_id", "default_user")
             if not message:
                 await websocket.send_json({"error": "missing message"})
@@ -435,4 +435,10 @@ if __name__ == "__main__":
     import uvicorn
     # 使用环境变量或默认端口 8079（避免与 llama.cpp 的 8081 端口冲突）
     port = int(BACKEND_PORT)
-    uvicorn.run(app, host="0.0.0.0", port=port)
+    uvicorn.run(
+        app, 
+        host="0.0.0.0", 
+        port=port,
+        log_level="debug",
+        access_log=True
+    )