app.py

from transformers import pipeline
import gradio as gr

# Load your model with pipeline
pipe = pipeline("text-generation", model="prithivMLmods/rStar-Coder-Qwen3-0.6B")

# Keep a conversation history
history = []

def chat_fn(user_input):
    global history
    # Add user message
    history.append(f"User: {user_input}")
    
    # Combine all previous messages as context
    context = "\n".join(history) + "\nBot:"
    
    # Generate response
    output = pipe(context, max_new_tokens=200, do_sample=True, top_p=0.9)[0]['generated_text']
    
    # Extract only the bot's reply (after last "Bot:")
    bot_reply = output.split("Bot:")[-1].strip()
    
    # Add bot reply to history
    history.append(f"Bot: {bot_reply}")
    
    return bot_reply

# Gradio interface
with gr.Blocks() as demo:
    chatbot_ui = gr.Chatbot()
    msg = gr.Textbox(placeholder="Type a message...")
    
    def respond(user_input, chat_history):
        reply = chat_fn(user_input)
        chat_history.append((user_input, reply))
        return chat_history, chat_history
    
    state = gr.State([])  # store chat history in Gradio
    msg.submit(respond, [msg, state], [chatbot_ui, state])

demo.launch()