Update app.py

app.py

@@ -38,15 +38,16 @@ def respond(message, history: list[tuple[str, str]], system_message, max_tokens,
     # If streaming is not supported, you can simply do:
     #   complete_response = agent.chat_completion(messages, max_tokens, temperature, top_p)
     #   yield complete_response
-    for token in agent.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        temperature=temperature,
-        top_p=top_p,
-        stream=True  # set to False if your agent does not support streaming
-    ):
-        complete_response += token
-        yield complete_response
+    # for token in agent.chat_completion(
+    #     messages,
+    #     max_tokens=max_tokens,
+    #     temperature=temperature,
+    #     top_p=top_p,
+    #     stream=True  # set to False if your agent does not support streaming
+    # ):
+    #     complete_response += token
+    complete_response=agent.run(messages)
+    yield complete_response
 
 # Step 3: Create the Gradio ChatInterface.
 demo = gr.ChatInterface(