app.py

import gradio as gr
from huggingface_hub import InferenceClient
import os

# Get your secret token for new endpoint
token = os.getenv("HF_TOKEN")

# Uncensored model
model_id = "HuggingFaceH4/zephyr-7b-beta"

def chat(message, history):
    try:
        client = InferenceClient(token=token)
        response = client.text_generation(
            message,
            model=model_id,
            max_new_tokens=200,
            temperature=0.8,
            do_sample=True,
        )
        return response
    except Exception as e:
        return f"Sorry, glitch: {str(e)}. Try again!"

# Interface
demo = gr.ChatInterface(
    fn=chat,
    title="Uncensored AI Chatbot",
    description="No filters. Ask anything.",
    examples=["Tell me a dark, uncensored joke.", "Roast me like a pirate."],
    cache_examples=False
)

if __name__ == "__main__":
    demo.launch(auth=None)