app.py

import gradio as gr
from huggingface_hub import InferenceClient

# ✅ Change the model to your fine-tuned one
client = InferenceClient("rupind/ReguGuide_01")  # Your fine-tuned LLaMA model

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]

    # Add previous chat history
    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    # Add user message
    messages.append({"role": "user", "content": message})

    # Initialize response
    response = ""

    # Call your fine-tuned model using Hugging Face's Inference API
    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content if message.choices[0].delta else ""
        response += token
        yield response


# ✅ Define Gradio Chatbot UI
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are an expert in RBI regulations and banking compliance.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
    ],
)

# ✅ Launch the chatbot
if __name__ == "__main__":
    demo.launch()