app.py

import gradio as gr

# This uses the Hugging Face Inference API (Serverless) 
# It is the best way to run large models on the Free CPU tier.
title = "Qwen 3.5 9B Uncensored Chat"
description = "A private instance of the Qwen 3.5 9B model."

interface = gr.load(
    name="models/LEONW24/Qwen3.5-9B-Uncensored",
    title=title,
    description=description,
    src="hf" # Tells Gradio to use the Hugging Face API
)

interface.launch()