Update app.py

app.py

@@ -15,9 +15,16 @@ model = AutoModelForCausalLM.from_pretrained(
     trust_remote_code=True
 )
 
-def chat(user_input, history):
-    # history is already in messages format
-    messages = history + [{"role": "user", "content": user_input}]
+SYSTEM_PROMPT = {
+    "role": "system",
+    "content": "You are a helpful AI assistant."
+}
+
+def chat(user_input, chat_history, message_state):
+    # message_state stores Qwen-style messages
+    messages = message_state or [SYSTEM_PROMPT]
+
+    messages.append({"role": "user", "content": user_input})
 
     prompt = tokenizer.apply_chat_template(
         messages,
@@ -42,16 +49,28 @@ def chat(user_input, history):
 
     messages.append({"role": "assistant", "content": response})
 
-    return messages, ""
+    chat_history.append((user_input, response))
+
+    return chat_history, messages, ""
 
 with gr.Blocks() as demo:
     gr.Markdown("## 🤖 Qwen Chatbot")
 
-    chatbot = gr.Chatbot(type="messages")
+    chatbot = gr.Chatbot()
     msg = gr.Textbox(label="Your message", autofocus=True)
+    state = gr.State([])  # stores Qwen messages
     clear = gr.Button("Clear")
 
-    msg.submit(chat, [msg, chatbot], [chatbot, msg])
-    clear.click(lambda: [], None, chatbot)
+    msg.submit(
+        chat,
+        inputs=[msg, chatbot, state],
+        outputs=[chatbot, state, msg]
+    )
+
+    clear.click(
+        lambda: ([], []),
+        None,
+        outputs=[chatbot, state]
+    )
 
 demo.launch()