app.py

import gradio as gr
from huggingface_hub import InferenceClient

client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

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

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

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

    response = ""

    for response_chunk in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = response_chunk.choices[0].delta.content
        response += token

    # Implement a basic check for relevance
    if not is_constitution_related(response):
        response = "Sorry, I can only answer questions related to the Constitution of India."

    yield response

def is_constitution_related(response):
    # Perform a simple check to see if the response seems related to the Constitution
    # This can be improved based on specific needs and feedback
    return "constitution" in response.lower() or "article" in response.lower()

demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a knowledgeable assistant specializing in the Constitution of India. Only provide answers related to the Constitution. If the question is not related, inform the user accordingly.", 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)",
        ),
    ],
)

if __name__ == "__main__":
    demo.launch()