Update app.py

app.py

@@ -132,32 +132,43 @@ async def access_control(request: Request, call_next):
             
     return await call_next(request)
 
-# 流式响应生成器
+# 修改 stream_generator 函数
 async def stream_generator(response):
+    buffer = ""
+    first_chunk = True
     try:
         async for chunk in response.aiter_bytes():
+            if first_chunk:
+                # 第一个响应来得快一些
+                await asyncio.sleep(0.1)
+                first_chunk = False
+            else:
+                # 后续响应平滑输出
+                await asyncio.sleep(0.05)
+                
             chunk_str = chunk.decode('utf-8')
-            for line in chunk_str.split('\n'):
+            buffer += chunk_str
+            
+            while '\n' in buffer:
+                line, buffer = buffer.split('\n', 1)
                 line = line.strip()
                 if line:
                     if line.startswith('data: '):
                         yield f"{line}\n\n"
                     else:
                         yield f"data: {line}\n\n"
-            # 添加小延迟以保持连接
-            await asyncio.sleep(0.01)
+                        
     except Exception as e:
         print(f"Stream error: {str(e)}")
         yield f"data: [ERROR] {str(e)}\n\n"
 
-# 处理API请求
+# 修改 handle_api_request 函数中的超时设置
 async def handle_api_request(url: str, headers: dict, method: str = "GET", body: dict = None, for_chat: bool = False):
     max_retries = 3
     current_try = 0
     
     while current_try < max_retries:
         try:
-            # 获取key
             if for_chat:
                 key = get_chat_key()
             else:
@@ -168,8 +179,15 @@ async def handle_api_request(url: str, headers: dict, method: str = "GET", body:
                 
             headers["Authorization"] = f"Bearer {key}"
             
-            # 修改超时设置
-            async with httpx.AsyncClient(timeout=httpx.Timeout(30.0, read=None)) as client:
+            # 减少连接超时，增加读取超时
+            timeout = httpx.Timeout(
+                connect=5.0,    # 连接超时5秒
+                read=60.0,      # 读取超时60秒
+                write=5.0,      # 写入超时5秒
+                pool=5.0        # 连接池超时5秒
+            )
+            
+            async with httpx.AsyncClient(timeout=timeout) as client:
                 response = await client.request(
                     method=method,
                     url=url,