Remove type="messages" from ChatInterface - use compatible Gradio API

app.py

@@ -29,18 +29,16 @@ def load_model():
 def respond(message, history, system_message, max_new_tokens, temperature, top_p, enable_thinking):
     load_model()
 
-    # Build messages list
-    # history is a list of {"role": ..., "content": ...} dicts in gr.ChatInterface
+    # Build messages list - handle both tuple and dict history formats
     messages = [{"role": "system", "content": system_message}]
     for item in history:
         if isinstance(item, dict):
             messages.append({"role": item["role"], "content": item["content"]})
-        else:
-            # fallback for tuple format
+        elif isinstance(item, (list, tuple)) and len(item) == 2:
             if item[0]:
-                messages.append({"role": "user", "content": item[0]})
+                messages.append({"role": "user", "content": str(item[0])})
             if item[1]:
-                messages.append({"role": "assistant", "content": item[1]})
+                messages.append({"role": "assistant", "content": str(item[1])})
     messages.append({"role": "user", "content": message})
 
     text = tokenizer.apply_chat_template(
@@ -89,7 +87,7 @@ def respond(message, history, system_message, max_new_tokens, temperature, top_p
 demo = gr.ChatInterface(
     fn=respond,
     title="Qwen3.5-35B-A3B AWQ Chat",
-    description="Powered by ZeroGPU (H200) | 4-bit AWQ quantized | 25.5 GB\n\n> **Note:** First inference takes ~2 min to load the model. Subsequent ones are faster.",
+    description="Powered by ZeroGPU (H200) | 4-bit AWQ quantized | 25.5 GB\n\n**Note:** First inference takes ~2 min to load the model. Subsequent ones are faster.",
     additional_inputs=[
         gr.Textbox(
             value="You are a helpful, smart, and concise AI assistant.",
@@ -107,7 +105,6 @@ demo = gr.ChatInterface(
         ["Explain quantum entanglement in simple terms."],
         ["Translate 'Hello, how are you?' into French, Spanish, and Japanese."],
     ],
-    type="messages",
 )
 
 demo.launch()