recovery/scripts/prepare_bindingdb_benchmark.py

#!/usr/bin/env python3
"""CLI wrapper for :func:`mitointeract_recovery.bindingdb_benchmark.prepare_benchmark`.

Builds the assay- and citation-stratified BindingDB exact-Kd benchmark
(``sample.jsonl``, four pair-level split manifests, deterministic
``audit.json``) from the pinned ``gold_exact_kd.jsonl`` and its source
audit. All validation, aggregation, and split logic lives in
``mitointeract_recovery.bindingdb_benchmark``; this script only parses
arguments and prints the resulting audit.
"""

from __future__ import annotations

import argparse
import json
from pathlib import Path

from mitointeract_recovery.bindingdb_benchmark import prepare_benchmark


def main() -> None:
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument(
        "--gold-jsonl",
        type=Path,
        required=True,
        help="path to gold_exact_kd.jsonl from prepare_bindingdb_gold_kd.py",
    )
    parser.add_argument(
        "--source-audit",
        type=Path,
        required=True,
        help="path to bindingdb_audit.json pinning the gold file checksums",
    )
    parser.add_argument("--output-dir", type=Path, required=True)
    parser.add_argument("--seed", type=int, default=42)
    args = parser.parse_args()

    audit = prepare_benchmark(
        gold_jsonl=args.gold_jsonl,
        source_audit=args.source_audit,
        output_dir=args.output_dir,
        seed=args.seed,
    )
    print(json.dumps(audit, indent=2, sort_keys=True))


if __name__ == "__main__":
    main()