app.py

import os
import gradio as gr
from huggingface_hub import InferenceClient

# Lee el token desde los secretos de tu Space (Settings → Repository secrets → HF_TOKEN)
HF_TOKEN = os.getenv("HF_TOKEN")

# Modelo base (usa Mistral oficial)
MODEL = "mistralai/Mistral-7B-Instruct-v0.3"

client = InferenceClient(model=MODEL, token=HF_TOKEN)

def chat_with_mistral(message):
    try:
        response = client.text_generation(
            prompt=message,
            max_new_tokens=200,
            temperature=0.7,
            repetition_penalty=1.1,
        )
        return response
    except Exception as e:
        return f"⚠️ Error: {str(e)}"

iface = gr.Interface(
    fn=chat_with_mistral,
    inputs="text",
    outputs="text",
    title="💬 Mistral Replica Chat",
    description="Chat de prueba usando el modelo Mistral desde Hugging Face API",
)

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