feat: support stream=false

main.py

@@ -65,6 +65,8 @@ async def get_models():
 async def generate_openai_chat_completion(form_data: OpenAIChatCompletionForm):
     user_message = get_last_user_message(form_data.messages)
 
+    if form_data.stream:
+
         def stream_content():
             res = get_response(user_message, messages=form_data.messages)
 
@@ -91,6 +93,36 @@ async def generate_openai_chat_completion(form_data: OpenAIChatCompletionForm):
             yield f"data: [DONE]"
 
         return StreamingResponse(stream_content(), media_type="text/event-stream")
+    else:
+
+        res = get_response(user_message, messages=form_data.messages)
+
+        message = ""
+
+        if isinstance(res, str):
+            message = stream_message_template(res)
+
+        elif isinstance(res, Generator):
+            for stream in res:
+                message = f"{message}{stream}"
+
+        return {
+            "id": f"rag-{str(uuid.uuid4())}",
+            "object": "chat.completion",
+            "created": int(time.time()),
+            "model": MODEL_ID,
+            "choices": [
+                {
+                    "index": 0,
+                    "message": {
+                        "role": "assistant",
+                        "content": message,
+                    },
+                    "logprobs": None,
+                    "finish_reason": "stop",
+                }
+            ],
+        }
 
 
 @app.get("/")

schemas.py

@@ -10,6 +10,7 @@ class OpenAIChatMessage(BaseModel):
 
 
 class OpenAIChatCompletionForm(BaseModel):
+    stream: bool = True
     model: str
     messages: List[OpenAIChatMessage]