Update main.py

main.py

@@ -32,11 +32,22 @@ async def index():
 class ChatCompletionRequest(BaseModel):
     prompt: str
 
+@app.get("/stream")
+async def chat(prompt = "Once upon a time there was a "):
+    completion = llm(prompt)
+    async def server_sent_events(chat_chunks):
+        for chat_chunk in chat_chunks:
+            yield dict(data=json.dumps(chat_chunk))
+        yield dict(data="[DONE]")
+
+    return EventSourceResponse(server_sent_events(completion))
+
 @app.post("/v1/chat/completions")
 async def chat(request: ChatCompletionRequest, response_mode=None):
     completion = llm(request.prompt)
     async def server_sent_events(chat_chunks):
         for chat_chunk in chat_chunks:
+            print(chat_chunk)
             yield dict(data=json.dumps(chat_chunk))
         yield dict(data="[DONE]")