Fix: use Gradio messages format (type="messages") to resolve data incompatibility error

app.py

@@ -29,12 +29,10 @@ def load_model():
 def respond(message, history, system_message, max_new_tokens, temperature, top_p, enable_thinking):
     load_model()
 
+    # Build messages list from history (new Gradio messages format)
     messages = [{"role": "system", "content": system_message}]
-    for user_msg, assistant_msg in history:
-        if user_msg:
-            messages.append({"role": "user", "content": user_msg})
-        if assistant_msg:
-            messages.append({"role": "assistant", "content": assistant_msg})
+    for item in history:
+        messages.append({"role": item["role"], "content": item["content"]})
     messages.append({"role": "user", "content": message})
 
     text = tokenizer.apply_chat_template(
@@ -83,7 +81,7 @@ def respond(message, history, system_message, max_new_tokens, temperature, top_p
 with gr.Blocks(title="Qwen3.5-35B-A3B Chat") as demo:
     gr.Markdown("# Qwen3.5-35B-A3B AWQ Chat\nPowered by ZeroGPU (H200) | 4-bit AWQ quantized | 25.5 GB\n\n> Note: First inference will take ~2 min to load the model. Subsequent ones are faster.")
 
-    chatbot = gr.Chatbot(height=500, label="Qwen3.5-35B-A3B")
+    chatbot = gr.Chatbot(height=500, label="Qwen3.5-35B-A3B", type="messages")
 
     with gr.Row():
         msg = gr.Textbox(
@@ -104,13 +102,17 @@ with gr.Blocks(title="Qwen3.5-35B-A3B Chat") as demo:
         top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p")
 
     def user_turn(user_message, history):
-        return "", history + [[user_message, None]]
+        history = history or []
+        history.append({"role": "user", "content": user_message})
+        history.append({"role": "assistant", "content": ""})
+        return "", history
 
     def bot_turn(history, system_message, max_new_tokens, temperature, top_p, enable_thinking):
-        user_message = history[-1][0]
-        bot_history = history[:-1]
-        for chunk in respond(user_message, bot_history, system_message, max_new_tokens, temperature, top_p, enable_thinking):
-            history[-1][1] = chunk
+        # history[-1] is the empty assistant placeholder, history[-2] is the user message
+        user_message = history[-2]["content"]
+        prior_history = history[:-2]
+        for chunk in respond(user_message, prior_history, system_message, max_new_tokens, temperature, top_p, enable_thinking):
+            history[-1]["content"] = chunk
             yield history
 
     msg.submit(user_turn, [msg, chatbot], [msg, chatbot]).then(