init inference

app.py

@@ -21,19 +21,23 @@ def chat_with_model(user_input, chat_history):
     :param chat_history: List of [user_message, ai_response] pairs
     :return: Updated chat history
     """
-    # Combine chat history into a single prompt
+    # Construct the prompt from chat history
     prompt = ""
     for user, ai in chat_history:
         prompt += f"User: {user}\nAI: {ai}\n"
-    prompt += f"User: {user_input}\nAI:"
+    prompt += f"User: {user_input}\nAI:"  # Add the latest user input
 
     # Generate response from the model
-    response = llm(prompt)["choices"][0]["text"].strip()
+    raw_response = llm(prompt)["choices"][0]["text"].strip()
 
-    # Update chat history as a list of tuples
+    # Clean the response (remove extra tags, if any)
+    response = raw_response.split("User:")[0].strip()
+
+    # Update chat history with the new turn
     chat_history.append((user_input, response))
     return chat_history, chat_history
 
+
 # Gradio UI
 with gr.Blocks() as demo:
     gr.Markdown("# 🦙 LLaMA GGUF Chatbot")