app.py

"""axentx coder-zero-gpu — thin proxy to HF Inference Router.

Why simple gradio? Previous attempts mixed FastAPI + gradio mount which
crashed on zero-a10g startup. This version is pure gradio app — gradio
auto-exposes a Predict API at /api/predict that we call from the LLM
chain. ZeroGPU not actually used (we don't load a model — just proxy)
so any tier works.
"""
import json
import os
import urllib.request
import gradio as gr

HF_TOKEN = os.environ.get("HF_TOKEN", "")
MODEL = os.environ.get("MODEL_ID", "Qwen/Qwen3-Coder-30B-A3B-Instruct")


def chat(prompt: str, system: str = "", max_tokens: int = 1024,
         temperature: float = 0.3) -> str:
    messages = []
    if system:
        messages.append({"role": "system", "content": system})
    messages.append({"role": "user", "content": prompt})
    body = json.dumps({
        "model": MODEL,
        "messages": messages,
        "max_tokens": int(max_tokens),
        "temperature": float(temperature),
    }).encode()
    req = urllib.request.Request(
        "https://router.huggingface.co/v1/chat/completions",
        data=body, method="POST",
        headers={"Authorization": f"Bearer {HF_TOKEN}",
                 "Content-Type": "application/json"})
    try:
        with urllib.request.urlopen(req, timeout=60) as r:
            d = json.loads(r.read())
        return d["choices"][0]["message"]["content"]
    except Exception as e:
        return f"ERROR: {type(e).__name__}: {e}"


# Simple gradio app — Space's auto-exposed API at /api/predict accepts
# {data: [prompt, system, max_tokens, temperature]} and returns {data:[output]}
demo = gr.Interface(
    fn=chat,
    inputs=[
        gr.Textbox(label="prompt", lines=4),
        gr.Textbox(label="system", lines=2, value=""),
        gr.Number(label="max_tokens", value=1024),
        gr.Number(label="temperature", value=0.3),
    ],
    outputs=gr.Textbox(label="output", lines=10),
    title="axentx coder zero-gpu",
    description="Thin proxy to HF Router. Use /api/predict from LLM chain.",
)

if __name__ == "__main__":
    demo.queue().launch(server_name="0.0.0.0", server_port=7860)