3.1 KiB
Scripts
This folder contains simple local scripts for validation and selector demos.
Current script:
validate_model_library.pycheck_card_contract.pyrebuild_indexes.pycheck_model_card_sync.pyrun_selector_regression.pyrun_selector_demo.pyrun_selector_calibration_smoke.pyrun_round04_blind_routing.pyrun_round04_post_patch_verification.pyinit_round05_context.pyrun_round05_1_selector_patch_audit.py
Run:
python scripts\validate_model_library.py
The script checks JSON parsing, required model fields, unique IDs, source references, excerpt references, regression case model references, and model/card index drift. It writes reports/validation_report.md.
Rebuild or check model/card indexes:
python scripts\rebuild_indexes.py --write
python scripts\rebuild_indexes.py --check
The index script regenerates models/model_index.json and cards/card_index.md from model JSON, card files, and regression cases. It writes reports/index_rebuild_report.md.
Run the card contract checker:
python scripts\check_card_contract.py
Run the selector demo:
python scripts\run_selector_demo.py
Run selector regression and model/card sync checks:
python scripts\run_selector_regression.py
python scripts\run_selector_calibration_smoke.py
python scripts\check_model_card_sync.py
run_selector_calibration_smoke.py checks selector calibration inputs, including strict no_call, no_call_or_low_priority, select_qpi_reject_ia, and IA heavy-depth gate expectations.
Run Round 04 blind routing evaluation:
python scripts\run_round04_blind_routing.py
This reads selector/round04_blind_inputs.json and writes reports/Round04_blind_routing_evaluation_report_2026-06-18.md. It does not score correctness or use expected routing labels.
Run Round 04.1 post-patch routing verification:
python scripts\run_round04_post_patch_verification.py
This reruns the frozen Round 04 input pool after selector rule patches and writes reports/Round04_1_post_patch_routing_verification_report_2026-06-18.md. It is not a second blind test.
Initialize Round 05 new-session context:
python scripts\init_round05_context.py
This reads Round 04 / 04.1 reports, selector source files, regression files, and workflow decisions, then writes ccra_review_bundle/round-05_init_selector-calibration-policy/00_ROUND05_NEW_SESSION_CONTEXT.md. It does not modify selector rules, model cards, model specs, or regression cases.
Run Round 05.1 selector patch audit:
python scripts\run_round05_1_selector_patch_audit.py baseline
python scripts\run_round05_1_selector_patch_audit.py after
python scripts\run_round05_1_selector_patch_audit.py diff
This captures before / after selector behavior for the Round 04 blind pool, selector calibration inputs, and aggregate regression cases. The diff report includes selected models, rejected models, scores, no-call status, reasons, penalties, target classification, changed fields, and non-target disposition placeholders.
Prefer Python standard library before adding dependencies.