Merge branch 'dev_fastchat_me' into dev_fastchat

server/chat/openai_chat.py

@@ -3,26 +3,40 @@ from fastapi.responses import StreamingResponse
 from typing import List, Dict
 import openai
 from configs.model_config import llm_model_dict, LLM_MODEL
+from pydantic import BaseModel
 
-async def openai_chat(messages: List[Dict] = Body(...,
+
-                                                  description="用户输入",
+class OpenAiMessage(BaseModel):
-                                                  example=[{"role": "user", "content": "你好"}])):
+    role: str = "user"
+    content: str = "hello"
+
+
+class OpenAiChatMsgIn(BaseModel):
+    model: str = LLM_MODEL
+    messages: List[OpenAiMessage]
+    temperature: float = 0.7
+    n: int = 1
+    max_tokens: int = 1024
+    stop: List[str] = []
+    stream: bool = True
+    presence_penalty: int = 0
+    frequency_penalty: int = 0
+
+
+async def openai_chat(msg: OpenAiChatMsgIn):
     openai.api_key = llm_model_dict[LLM_MODEL]["api_key"]
     print(f"{openai.api_key=}")
     openai.api_base = llm_model_dict[LLM_MODEL]["api_base_url"]
     print(f"{openai.api_base=}")
-    print(messages)
+    print(msg)
 
-    async def get_response(messages: List[Dict]):
+    async def get_response(msg):
-        response = openai.ChatCompletion.create(
+        response = openai.ChatCompletion.create(**msg.dict())
-            model=LLM_MODEL,
-            messages=messages,
-        )
         for chunk in response.choices[0].message.content:
             print(chunk)
             yield chunk
 
     return StreamingResponse(
-        get_response(messages),
+        get_response(msg),
         media_type='text/event-stream',
-    )
+    )