app.py

import os
import gradio as gr
from huggingface_hub import InferenceClient

# --- Configuration ---
# Connect directly via Hugging Face's Native Client
hf_token = os.environ.get("HF_TOKEN")
if not hf_token:
    raise ValueError("HF_TOKEN missing. Please add your Hugging Face Access Token to the Space Secrets.")

# Initialize the Zephyr model
client = InferenceClient(
    model="Qwen/Qwen2.5-7B-Instruct",
    token=hf_token
)

def chat_function(message, history):
    """
    This function takes the user's message and the chat history,
    formats it for the Hugging Face API, and returns the response.
    """
    # 1. Format the conversation history into the required structure
    messages = []
    for user_msg, ai_msg in history:
        messages.append({"role": "user", "content": user_msg})
        messages.append({"role": "assistant", "content": ai_msg})
    
    # 2. Add the current user's message
    messages.append({"role": "user", "content": message})

    # 3. Call the model
    try:
        response = client.chat_completion(messages, max_tokens=512)
        return response.choices[0].message.content
    except Exception as e:
        return f"An error occurred: {str(e)}"

# --- Gradio UI ---
# Use Gradio's built-in ChatInterface for a clean look
demo = gr.ChatInterface(
    fn=chat_function,
    title="Direct Chat with Qwen/Qwen2.5-7B-Instruct",
    description="This app talks directly to the Qwen/Qwen2.5-7B-Instruct model via Hugging Face Serverless API. No local database is attached."
)

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