import gradio as gr
from huggingface_hub import InferenceApi

# point to the base model with no instruction tuning or filters
api = InferenceApi(
    repo_id="EleutherAI/gpt-neo-1.3B",
    task="text-generation"
)

def respond(prompt):
    try:
        outputs = api(inputs=prompt, parameters={"max_new_tokens":200, "temperature":0.7})
        return outputs[0].get("generated_text", "[error] no text returned")
    except Exception as e:
        return f"[exception] {e}"

gr.Interface(
    fn=respond,
    inputs=gr.Textbox(lines=3, placeholder="type anything…"),
    outputs="text",
    title="uncensored gpt-neo-1.3B chat"
).launch(server_name="0.0.0.0", server_port=7860)