Update app.py

app.py

@@ -29,7 +29,7 @@ async def respond(
 
     response = ""
 
-    for message in client.chat_completion(
+    async for message in client.chat_completion(
         messages,
         max_tokens=max_tokens,
         stream=True,
@@ -40,13 +40,15 @@ async def respond(
 
         response += token
         yield response
+        
 
-
+def sync_respond(*args):
+    return asyncio.run(respond(*args))
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
 """
 demo = gr.ChatInterface(
-    respond,
+    sync_respond,
     additional_inputs=[
         gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),