Update app.py

app.py

@@ -81,12 +81,15 @@ def predict_chat(message: str, history: list):
         yield "Error: Model or tokenizer failed to load. Please check the Space logs for details."
         return
 
-    # Gradio history is already formatted as a list of lists: [[user_msg, bot_msg], ...]
-    # We need to convert it to the format expected by the tokenizer's chat template.
+    # Initialize messages list with system message
     messages = [{"role": "system", "content": "You are a friendly chatbot."}]
-    for human, assistant in history:
-        messages.append({"role": "user", "content": human})
-        messages.append({"role": "assistant", "content": assistant})
+    
+    # Extend messages with the existing history directly
+    # Gradio's gr.Chatbot(type='messages') passes history as a list of dictionaries 
+    # with 'role' and 'content' keys, which is compatible with apply_chat_template.
+    messages.extend(history)
+
+    # Append the current user message
     messages.append({"role": "user", "content": message})
 
     generated_text = ""
@@ -97,7 +100,6 @@ def predict_chat(message: str, history: list):
         print("Using GGUF model generation path.")
         # Apply chat template for GGUF models as well,
         # though ctransformers might expect a simpler string.
-        # This can be adjusted if the model has a specific prompt format.
         # For Llama-based models, the tokenizer.apply_chat_template should work.
         prompt_input = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)