app.py

import gradio as gr
import random 
from huggingface_hub import InferenceClient

client = InferenceClient("Qwen/Qwen2.5-7B-Instruct")

def respond(message, history): 
    messages = [{"role" : "system", "content" : "you are a friendly smth"}]

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

    response = client.chat_completion(
        messages,
        max_tokens = 100
    )
    return response.choices[0].messages.content.strip()
    
chatbot = gr.ChatInterface(respond)
chatbot.launch()