app.py

import gradio as gr
from huggingface_hub import InferenceClient
import os

system_message = os.environ["SYSTEM_MESSAGE"]
HF_TOKEN = os.environ["HF_TOKEN"]
MODEL_NAME = os.environ["MODEL_NAME"]

client = InferenceClient(token=HF_TOKEN)

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

    
    for user_msg, assistant_msg in history:
        prompt.append({"role": "user", "content": user_msg})
        prompt.append({"role": "assistant", "content": assistant_msg})

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

    response = []
    stream = client.chat.completions.create(
        model=MODEL_NAME,
        messages=prompt,
        max_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        stream=True
    )

    for chunk in stream:
        if not chunk.choices:
            continue
        delta = chunk.choices[0].delta
        token = getattr(delta, "content", None)
        if token:
            response.append(token)
            yield "".join(response)


app = gr.ChatInterface(
    fn=respond,
    additional_inputs=[
        gr.Slider(16, 2048, value=512, step=1, label="Max Tokens"),
        gr.Slider(0.1, 2.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p"),
    ],
)

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