Function main

eval/compare_results.py:202–226

(argv: list[str] | None = None)

def main(argv: list[str] | None = None) -> int:
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("--left", required=True, type=Path, help="Path to first summary JSON.")
    parser.add_argument("--right", required=True, type=Path, help="Path to second summary JSON.")
    parser.add_argument("--left-label", default="left", help="Label for the first agent.")
    parser.add_argument("--right-label", default="right", help="Label for the second agent.")
    parser.add_argument(
        "--out",
        type=Path,
        default=None,
        help="Optional path to write the markdown report. Disagreement lists are written next to it.",
    )
    args = parser.parse_args(argv)

    left = load_summary(args.left, args.left_label)
    right = load_summary(args.right, args.right_label)

    print(render_text(left, right))

    if args.out is not None:
        args.out.parent.mkdir(parents=True, exist_ok=True)
        args.out.write_text(render_markdown(left, right), encoding="utf-8")
        write_disagreement_lists(args.out.parent, left, right)
        print(f"\nWrote {args.out}")
    return 0


if __name__ == "__main__":