from __future__ import annotations

from core.device_manager import DeviceManager
from core.models import Scene
from tools._device import get_driver
from vision.ocr import PaddleOCREngine, run_ocr
from vision.scene_builder import build_scene, infer_png_size
from vision.ui_parser import parse_ui_tree


def describe_screen(
    device_id: str | None = None,
    *,
    manager: DeviceManager | None = None,
    ocr_engine: PaddleOCREngine | None = None,
) -> Scene:
    driver = get_driver(device_id, manager=manager)
    screenshot = driver.screenshot()
    raw_tree = driver.tree()
    width, height = infer_png_size(screenshot)
    return build_scene(
        screen_width=width,
        screen_height=height,
        ui_elements=parse_ui_tree(raw_tree),
        ocr_elements=run_ocr(screenshot, engine=ocr_engine),
    )