update

app.py

@@ -16,17 +16,22 @@ model = AutoModelForCausalLM.from_pretrained(
 )
 
 # 2) define inference function
-def generate(messages):
+def generate(message, history):
     """
-    messages: List of alternating [user, assistant, user, ...]
-    returns: [user, assistant, user, assistant, ...] with model's reply appended
+    message: Current user message (string)
+    history: List of [user_message, assistant_message] pairs
+    returns: assistant's reply (string)
     """
-    # rebuild a single prompt string
+    # rebuild a single prompt string from history + current message
     prompt = ""
-    for i in range(0, len(messages), 2):
-        prompt += f"User: {messages[i]}\n"
-        if i+1 < len(messages):
-            prompt += f"Assistant: {messages[i+1]}\n"
+    
+    # Add conversation history
+    for user_msg, assistant_msg in history:
+        prompt += f"User: {user_msg}\n"
+        prompt += f"Assistant: {assistant_msg}\n"
+    
+    # Add current user message
+    prompt += f"User: {message}\n"
     prompt += "Assistant:"
 
     inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
@@ -40,8 +45,7 @@ def generate(messages):
     # strip everything before the last "Assistant:"
     reply = text.split("Assistant:")[-1].strip()
 
-    messages.append(reply)
-    return messages
+    return reply
 
 # 3) build Gradio ChatInterface
 demo = gr.ChatInterface(