app.py

import gradio as gr
from ctransformers import AutoModelForCausalLM

# Charger le modèle GGUF (léger)
model = AutoModelForCausalLM.from_pretrained(
    "TheBloke/TinyLlama-1.1B-Chat-v0.3-GGUF",
    model_file="TinyLlama-1.1B-Chat-v0.3.Q4_K_M.gguf",
    model_type="llama",
    gpu_layers=0
)

def chat(prompt):
    response = model(
        prompt,
        max_new_tokens=200,
        temperature=0.7,
        repetition_penalty=1.1
    )
    return response

with gr.Blocks() as demo:
    gr.Markdown("# API IA pour Discord (GGUF)")
    inp = gr.Textbox(label="Prompt")
    out = gr.Textbox(label="Réponse")
    btn = gr.Button("Envoyer")
    btn.click(chat, inp, out)

demo.launch(server_name="0.0.0.0", server_port=7860)