more

app.py

@@ -42,7 +42,9 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
         top_p=top_p,
     )
     response = tokenizer.decode(generated_tokens[0], skip_special_tokens=True)
-    yield [{"role": "assistant", "content": response}]
+
+    history.append((message, response))  # Update history with new exchange
+    return history  # Return the updated history
 
 # Define the Gradio interface
 demo = gr.ChatInterface(
@@ -57,4 +59,4 @@ demo = gr.ChatInterface(
 )
 
 if __name__ == "__main__":
-    demo.launch(share=True)
+    demo.launch(share=False)  # Set share=False for local testing