feat(server): 添加访问日志中间件并优化日志配置

1. 实现自定义访问日志中间件以记录请求处理时间
2. 禁用uvicorn默认访问日志处理器
优化日志格式化配置

server/main.py

@@ -12,6 +12,7 @@ from server.routers import router
 from server.utils.lifespan import lifespan
 from server.utils.auth_middleware import is_public_path
 from server.utils.common_utils import setup_logging
+from server.utils.access_log_middleware import AccessLogMiddleware
 
 # 设置日志配置
 setup_logging()
@@ -115,6 +116,9 @@ class AuthMiddleware(BaseHTTPMiddleware):
         return await call_next(request)
 
 
+# 添加访问日志中间件（记录请求处理时间）
+app.add_middleware(AccessLogMiddleware)
+
 # 添加鉴权中间件
 app.add_middleware(LoginRateLimitMiddleware)
 app.add_middleware(AuthMiddleware)

server/utils/access_log_middleware.py

@@ -0,0 +1,67 @@
+"""访问日志中间件 - 记录请求处理时间"""
+
+import time
+import logging
+from collections.abc import Callable
+
+from fastapi import Request, Response
+from starlette.middleware.base import BaseHTTPMiddleware
+
+# 创建专用的访问日志记录器
+access_logger = logging.getLogger("access_logger")
+
+# 设置访问日志记录器
+if not access_logger.handlers:
+    handler = logging.StreamHandler()
+    formatter = logging.Formatter(fmt="%(asctime)s %(levelname)s: %(message)s", datefmt="%m-%d %H:%M:%S")
+    handler.setFormatter(formatter)
+    access_logger.addHandler(handler)
+    access_logger.setLevel(logging.INFO)
+    # 避免传播到根日志记录器，防止重复日志
+    access_logger.propagate = False
+
+
+def _extract_client_ip(request: Request) -> str:
+    """提取客户端IP地址"""
+    forwarded_for = request.headers.get("x-forwarded-for")
+    if forwarded_for:
+        return forwarded_for.split(",")[0].strip()
+    if request.client:
+        return request.client.host
+    return "unknown"
+
+
+class AccessLogMiddleware(BaseHTTPMiddleware):
+    """访问日志中间件 - 记录请求处理时间"""
+
+    def __init__(self, app, logger: logging.Logger = None):
+        super().__init__(app)
+        self.logger = logger or access_logger
+
+    async def dispatch(self, request: Request, call_next: Callable) -> Response:
+        """处理请求并记录访问日志"""
+        # 记录请求开始时间
+        start_time = time.perf_counter()
+
+        # 获取客户端IP
+        client_ip = _extract_client_ip(request)
+
+        # 处理请求
+        response = await call_next(request)
+
+        # 计算处理时间
+        process_time = time.perf_counter() - start_time
+        process_time_ms = int(process_time * 1000)  # 转换为毫秒
+
+        # 格式化日志消息，添加处理时间
+        log_message = (
+            f"{client_ip}:{request.client.port if request.client else 'unknown'} - "
+            f'"{request.method} {request.url.path}{"?" + request.url.query if request.url.query else ""} '
+            f'HTTP/{request.scope["http_version"]}" '
+            f"{response.status_code} - {process_time_ms}ms"
+        )
+
+        # 记录日志
+        self.logger.info(log_message)
+
+        return response

server/utils/common_utils.py

@@ -19,14 +19,15 @@ def setup_logging():
     uvicorn_logger = logging.getLogger("uvicorn")
     uvicorn_access_logger = logging.getLogger("uvicorn.access")
 
+    # 禁用默认的uvicorn访问日志（因为我们使用自定义中间件）
+    uvicorn_access_logger.handlers.clear()
+
     # 创建格式化器
     formatter = logging.Formatter(fmt="%(asctime)s %(levelname)s: %(message)s", datefmt="%m-%d %H:%M:%S")
 
-    # 为所有处理器设置格式化器
+    # 为uvicorn主日志设置格式化器
     for handler in uvicorn_logger.handlers:
         handler.setFormatter(formatter)
-    for handler in uvicorn_access_logger.handlers:
-        handler.setFormatter(formatter)
 
 
 async def log_operation(db: Session, user_id: int, operation: str, details: str = None, request: Request = None):

src/agents/common/models.py

@@ -32,7 +32,6 @@ def load_chat_model(fully_specified_name: str, **kwargs) -> BaseChatModel:
         logger.debug(f"[offical] Loading model {model_spec} with kwargs {kwargs}")
         return init_chat_model(model_spec, **kwargs)
 
-
     elif provider in ["dashscope"]:
         from langchain_deepseek import ChatDeepSeek
 

src/agents/deep_agent/graph.py

@@ -1,14 +1,11 @@
 """Deep Agent - 基于create_deep_agent的深度分析智能体"""
 
-from langchain.agents.middleware import ModelRequest, dynamic_prompt, SummarizationMiddleware
-
-from langchain.agents import create_agent
-from langchain.agents.middleware import TodoListMiddleware
-from langchain_anthropic.middleware import AnthropicPromptCachingMiddleware
-
 from deepagents.middleware.filesystem import FilesystemMiddleware
 from deepagents.middleware.patch_tool_calls import PatchToolCallsMiddleware
 from deepagents.middleware.subagents import SubAgentMiddleware
+from langchain.agents import create_agent
+from langchain.agents.middleware import ModelRequest, SummarizationMiddleware, TodoListMiddleware, dynamic_prompt
+from langchain_anthropic.middleware import AnthropicPromptCachingMiddleware
 
 from src.agents.common import BaseAgent, load_chat_model
 from src.agents.common.middlewares import context_based_model, inject_attachment_context
@@ -21,9 +18,7 @@ search_tools = [search]
 
 research_sub_agent = {
     "name": "research-agent",
-    "description": (
+    "description": ("利用搜索工具，用于研究更深入的问题。"),
-        "利用搜索工具，用于研究更深入的问题。"
-    ),
     "system_prompt": (
         "你是一位专注的研究员。你的工作是根据用户的问题进行研究。"
         "进行彻底的研究，然后用详细的答案回复用户的问题，只有你的最终答案会被传递给用户。"

src/knowledge/implementations/lightrag.py

@@ -235,7 +235,7 @@ class LightRagKB(KnowledgeBase):
                 model=model_name,
                 api_key=config_dict["api_key"],
                 base_url=config_dict["base_url"].replace("/embeddings", ""),
-            )
+            ),
         )
 
     async def add_content(self, db_id: str, items: list[str], params: dict | None = None) -> list[dict]:

src/knowledge/utils/kb_utils.py

@@ -319,7 +319,7 @@ def get_embedding_config(embed_info: dict) -> dict:
             "model": embed_info["name"],
             "api_key": os.getenv(embed_info["api_key"]) or embed_info["api_key"],
             "base_url": embed_info["base_url"],
-            "dimension": embed_info.get("dimension", 1024)
+            "dimension": embed_info.get("dimension", 1024),
         }
         logger.debug(f"Embedding config from dict: {config_dict}")
         return config_dict