Update app.py

app.py

@@ -10,15 +10,18 @@ def respond(message, history):
         messages.extend(history)
     messages.append({"role": "user", "content": message})
 
-    response = client.chat_completion(
-        messages,
-        max_tokens=100,
-        temperature = 1.5,
-        top_p = 0.34
-    )
+    # response = client.chat_completion(
+    #     messages,
+    #     max_tokens=100,
+    #     temperature = 1.5,
+    #     top_p = 0.34
+    # )
+    response = client.chat_completion(stream=True)
+    for msg in response:
+        token = message.choices[0].delta.content
+        response += token
+        yield token #response.choices[0].message.content.strip()
     
-    return response.choices[0].message.content.strip()
-
 chatbot = gr.ChatInterface(respond, title="my bot")