"""Post one or more images to an OCR endpoint and save text_lines per image.

Used to freeze the accuracy baseline and to capture candidate outputs for CER.

Usage:
    python scripts/capture_ocr_text.py \
        --url http://127.0.0.1:5002/v1/api/ai/suya_ocr_vllm/ \
        --out-dir baseline_outputs \
        temp_image_c3c56948-282e-453b-8fac-6c482243d1e5.jpg
"""
from __future__ import annotations

import argparse
import base64
from pathlib import Path

import requests


def _payload(image_path: Path) -> dict:
    suffix = image_path.suffix.lstrip(".").lower() or "png"
    return {
        "file": base64.b64encode(image_path.read_bytes()).decode("utf-8"),
        "type": "jpg" if suffix == "jpeg" else suffix,
        "skip_text_detection": False,
        "skip_table_detection": False,
        "recognize_math": False,
        "ocr_with_boxes": True,
    }


def capture(url: str, image_path: Path, out_dir: Path, timeout: float) -> str:
    resp = requests.post(url, json=_payload(image_path), timeout=timeout)
    body = resp.json()
    text = (body.get("data") or {}).get("text_lines", "") if isinstance(body.get("data"), dict) else ""
    out_dir.mkdir(parents=True, exist_ok=True)
    (out_dir / (image_path.stem + ".txt")).write_text(text, encoding="utf-8")
    return text


def main() -> None:
    parser = argparse.ArgumentParser(description="Capture OCR text_lines per image.")
    parser.add_argument("images", nargs="+", type=Path)
    parser.add_argument("--url", required=True)
    parser.add_argument("--out-dir", type=Path, required=True)
    parser.add_argument("--timeout", type=float, default=900)
    args = parser.parse_args()
    for image in args.images:
        capture(args.url, image, args.out_dir, args.timeout)
        print(f"captured {image.name} -> {args.out_dir}")


if __name__ == "__main__":
    main()