"""tinydoc command line.

    python -m tinydoc extract receipt.jpg invoice.pdf --backend onnx --weights models
    python -m tinydoc params
"""

import argparse
import json
import sys


def main():
    ap = argparse.ArgumentParser(prog="tinydoc")
    sub = ap.add_subparsers(dest="cmd", required=True)
    ex = sub.add_parser("extract", help="extract structured JSON from images / PDFs")
    ex.add_argument("files", nargs="+")
    ex.add_argument("--backend", choices=["onnx", "torch"], default="onnx")
    ex.add_argument("--weights", default=None, help="models/ (onnx) or runs/ (torch)")
    ex.add_argument("--device", default="cpu")
    ex.add_argument("--force-ocr", action="store_true", help="OCR PDFs even when they contain usable text")
    ex.add_argument("--out", default=None, help="write JSON here (default: stdout)")
    ex.add_argument("--compact", action="store_true", help="omit per-segment detail")
    sub.add_parser("params", help="print the parameter budget")
    args = ap.parse_args()

    if args.cmd == "params":
        from .models import BUDGET, parameter_report
        for k, v in parameter_report().items():
            print(f"{k:12s} {v:>10,d}   (target {BUDGET.get(k, sum(BUDGET.values())):,d})")
        return

    from .pipeline.engine import Engine
    weights = args.weights or ("models" if args.backend == "onnx" else "runs")
    engine = Engine(weights, backend=args.backend, device=args.device)
    results = []
    for f in args.files:
        r = engine.process(f, force_ocr=args.force_ocr)
        if args.compact:
            for p in r["pages"]:
                p.pop("segments", None)
        results.append(r)
    payload = results[0] if len(results) == 1 else results
    text = json.dumps(payload, indent=2, ensure_ascii=False)
    if args.out:
        with open(args.out, "w") as fh:
            fh.write(text)
    else:
        sys.stdout.write(text + "\n")


if __name__ == "__main__":
    main()