from __future__ import annotations

import logging
import os
import threading

import spaces
from gradio import Server

from app.config import get_settings
from app.factory import create_app

logger = logging.getLogger(__name__)

GRADIO_DOC_ROUTES = {"/docs", "/openapi.json", "/redoc", "/docs/oauth2-redirect"}


@spaces.GPU(duration=1)
def _zero_gpu_probe() -> str:
    """Satisfy the ZeroGPU runtime requirement; the demo itself runs on CPU."""
    return "ok"


def _warm_whisper(transcriber) -> None:
    try:
        transcriber.warm_up()
        logger.info("Whisper model is loaded and ready")
    except Exception:
        logger.exception("Whisper warm-up failed; the first request will retry")


settings = get_settings()
api = create_app(settings)

# Gradio's Server owns the outer FastAPI instance. Mount the real application at
# the root so the dashboard, API, WebSocket, and docs are served unchanged.
demo = Server(title=settings.app_name)


@demo.get("/gradio_api/startup-events")
def _gradio_startup_events() -> dict[str, str]:
    # Gradio's launch() probes this path. The root mount below would otherwise
    # shadow Gradio's own handler, and the probe would fail with a 404.
    return {"status": "ok"}


demo.mount("/", api)
demo.routes[:] = [
    route for route in demo.routes if getattr(route, "path", None) not in GRADIO_DOC_ROUTES
]

if os.environ.get("SPACE_WARMUP", "1") == "1":
    threading.Thread(
        target=_warm_whisper,
        args=(api.state.transcriber,),
        name="whisper-warmup",
        daemon=True,
    ).start()

if __name__ == "__main__":
    demo.launch(ssr_mode=False)