Update app.py

app.py

@@ -74,47 +74,62 @@ def format_conversation(history, system_prompt):
     prompt += "Assistant: "
     return prompt
 
-def generate_response(user_input, history, system_prompt, model_name, max_tokens, temperature, top_k, top_p, repeat_penalty):
+# Function to get just the model name from the dropdown selection
+def get_model_name(full_selection):
+    return full_selection.split(" - ")[0]
+
+# User input handling function
+def user_input(user_message, history):
+    return "", history + [(user_message, None)]
+
+@spaces.GPU(duration=60)
+def bot_response(history, system_prompt, model_selection, max_tokens, temperature, top_k, top_p, repetition_penalty):
     """
-    Generate a complete response (non-streaming).
+    Generate AI response to user input
     """
     cancel_event.clear()
-    full_history = history.copy()
     
-    # Format conversation for the model
-    conversation = format_conversation(full_history, system_prompt)
+    # Extract the latest user message
+    user_message = history[-1][0]
+    history_without_last = history[:-1]
+    
+    # Get model name from selection
+    model_name = get_model_name(model_selection)
+    
+    # Format the conversation
+    conversation = format_conversation(history_without_last, system_prompt)
+    conversation += "User: " + user_message + "\nAssistant: "
     
     try:
         pipe = load_pipeline(model_name)
-        output = pipe(
+        response = pipe(
             conversation,
             max_new_tokens=max_tokens,
             temperature=temperature,
             top_k=top_k,
             top_p=top_p,
-            repetition_penalty=repeat_penalty,
+            repetition_penalty=repetition_penalty,
             return_full_text=False
         )[0]["generated_text"]
         
-        # Return the updated history
-        history.append((user_input, output))
+        # Update the last message pair with the response
+        history[-1] = (user_message, response)
         return history
     except Exception as e:
-        history.append((user_input, f"Error: {e}"))
+        history[-1] = (user_message, f"Error: {e}")
         return history
     finally:
         gc.collect()
 
-def cancel_generation():
-    cancel_event.set()
-    return 'Generation cancelled.'
-
 def get_default_system_prompt():
     today = datetime.now().strftime('%Y-%m-%d')
     return f"""You are Qwen3, a helpful and friendly AI assistant created by Alibaba Cloud.
 Today is {today}.
 Be concise, accurate, and helpful in your responses."""
 
+def clear_chat():
+    return []
+
 # CSS for improved visual style
 css = """
 .gradio-container {
@@ -162,10 +177,6 @@ css = """
 }
 """
 
-# Function to get just the model name from the dropdown selection
-def get_model_name(full_selection):
-    return full_selection.split(" - ")[0]
-
 # ------------------------------
 # Gradio UI
 # ------------------------------
@@ -217,51 +228,36 @@ with gr.Blocks(title="Qwen3 Chat", css=css) as demo:
     </div>
     """)
     
-    # Define event handlers
-    def user_input(user_message, history):
-        return "", history + [(user_message, None)]
-    
-    def bot_response(history, sys_prompt, model, max_tok, temp, k, p, rp):
-        user_message = history[-1][0]
-        bot_message = generate_response(
-            user_message,
-            history[:-1],
-            sys_prompt,
-            get_model_name(model),
-            max_tok,
-            temp,
-            k,
-            p,
-            rp
-        )[-1][1]
-        
-        history[-1] = (user_message, bot_message)
-        return history
-    
-    # Connect everything
+    # Connect UI elements to functions
     submit_btn.click(
         user_input,
-        [txt, chatbot],
-        [txt, chatbot],
+        inputs=[txt, chatbot],
+        outputs=[txt, chatbot],
         queue=False
     ).then(
         bot_response,
-        [chatbot, sys_prompt, model_dd, max_tok, temp, k, p, rp],
-        [chatbot]
+        inputs=[chatbot, sys_prompt, model_dd, max_tok, temp, k, p, rp],
+        outputs=chatbot,
+        api_name="generate"
     )
     
     txt.submit(
         user_input,
-        [txt, chatbot],
-        [txt, chatbot],
+        inputs=[txt, chatbot],
+        outputs=[txt, chatbot],
         queue=False
     ).then(
         bot_response,
-        [chatbot, sys_prompt, model_dd, max_tok, temp, k, p, rp],
-        [chatbot]
+        inputs=[chatbot, sys_prompt, model_dd, max_tok, temp, k, p, rp],
+        outputs=chatbot,
+        api_name="generate"
     )
     
-    clear_btn.click(lambda: None, None, chatbot, queue=False)
+    clear_btn.click(
+        clear_chat,
+        outputs=[chatbot],
+        queue=False
+    )
 
 if __name__ == "__main__":
     demo.launch()