Update app.py

app.py

@@ -11,13 +11,19 @@ def respond(message, history):
         messages.extend(history)
         
     messages.append({"role": "user", "content": message})
+
+    response = ""
     
-    response = client.chat_completion(
+    response_stream = client.chat_completion(
         messages,
-        max_tokens=100
+        max_tokens=100,
+        stream=True
     )
-    
-    return response['choices'][0]['message']['content'].strip()
+
+    for token in response_stream:
+        token = message['choices'][0]['message']['content'].strip()
+        response += token
+        yield response
 
 chatbot = gr.ChatInterface(respond, type="messages")