Update app.py

app.py

@@ -20,7 +20,7 @@ print("OpenAI client initialized.")
 
 def respond(
     message,
-    history: list[tuple[str, str]],
+    history,
     system_message,
     max_tokens,
     temperature,
@@ -53,9 +53,11 @@ def respond(
     # If user provided a model, use that; otherwise, fall back to a default model
     model_to_use = custom_model.strip() if custom_model.strip() != "" else "meta-llama/Llama-3.3-70B-Instruct"
     
-    # Start with an empty string to build the response as tokens stream in
-    response = ""
-
+    # Create a copy of the history and add the new user message
+    new_history = list(history)
+    new_history.append((message, ""))
+    current_response = ""
+    
     try:
         for message_chunk in client.chat.completions.create(
             model=model_to_use,
@@ -69,10 +71,14 @@ def respond(
         ):
             token_text = message_chunk.choices[0].delta.content
             if token_text is not None:  # Handle None type in response
-                response += token_text
-                yield response
+                current_response += token_text
+                # Update just the last message in history
+                new_history[-1] = (message, current_response)
+                yield new_history
     except Exception as e:
-        yield f"Error: {str(e)}\n\nPlease check your model selection and parameters, or try again later."
+        error_message = f"Error: {str(e)}\n\nPlease check your model selection and parameters, or try again later."
+        new_history[-1] = (message, error_message)
+        yield new_history
 
     print("Completed response generation.")
 
@@ -393,16 +399,24 @@ with gr.Blocks(css=custom_css, title=APP_TITLE, theme=gr.themes.Soft()) as demo:
         fn=respond,
         inputs=[msg, chatbot, system_message_box, max_tokens_slider, temperature_slider, 
                 top_p_slider, frequency_penalty_slider, seed_slider, custom_model_box],
-        outputs=[chatbot],
+        outputs=chatbot,
         queue=True
+    ).then(
+        lambda: "", # Clear the message box after sending
+        None,
+        [msg]
     )
     
     submit_btn.click(
         fn=respond,
         inputs=[msg, chatbot, system_message_box, max_tokens_slider, temperature_slider, 
                 top_p_slider, frequency_penalty_slider, seed_slider, custom_model_box],
-        outputs=[chatbot],
+        outputs=chatbot,
         queue=True
+    ).then(
+        lambda: "", # Clear the message box after sending
+        None,
+        [msg]
     )
     
     # Update model display when search changes