调整流式响应默认设置为非流式，确保与 OpenAI 兼容

src/api/routers/chat.py

@@ -207,12 +207,15 @@ async def create_chat_completion(request: ChatCompletionRequest):
     # Extract parameters
     params = extract_chat_params(request)
 
-    # Check global streaming setting
-    # If streaming is globally disabled, override client request
-    enable_stream = request.stream and settings.enable_streaming
+    # Determine streaming mode
+    # Priority: request.stream (client) > settings.enable_streaming (global)
+    # Only enable streaming if BOTH conditions are met:
+    # 1. Client explicitly requests stream=true (default is false per OpenAI standard)
+    # 2. Global streaming is enabled via ENABLE_STREAMING
+    enable_stream = request.stream is True and settings.enable_streaming
 
-    if not settings.enable_streaming and request.stream:
-        logger.info("Streaming requested but globally disabled via ENABLE_STREAMING=false")
+    if request.stream is True and not settings.enable_streaming:
+        logger.info("Streaming requested by client but globally disabled via ENABLE_STREAMING=false")
 
     # Handle streaming
     if enable_stream:

src/api/schemas.py

@@ -23,7 +23,7 @@ class ChatCompletionRequest(BaseModel):
     temperature: Optional[float] = 0.7
     top_p: Optional[float] = 1.0
     n: Optional[int] = 1
-    stream: Optional[bool] = True  # Default to streaming
+    stream: Optional[bool] = False  # Default to non-streaming (OpenAI compatible)
     stop: Optional[Union[str, List[str]]] = None
     max_tokens: Optional[int] = None
     presence_penalty: Optional[float] = 0.0