chore(fidelity): re-pin the table to the 51-language corpus (pre-v2.9.0) - #3187
Merged
Merged
Conversation
Regenerated gitgalaxy/standards/fidelity_table.py (+ the generated wiki tables) from keyword-rosetta main @ 83fd3487 -- the committed pin (6377eb59, 2026-09-04, 46 languages x 19 signals) predated the CICS wave (pli, bms, db2_sql, hlasm, rexx; rosetta #135-#138) and the units_public / units_documented signals. Now 51 languages x 21 signals, 33 cells below 1.0. The old sub-1.0 cells (branch/state_mutation/io/...) all cleared to exact-on-plant -- the alignment campaign's fixes, now certified by the table. New cells are dominated by units_public 12/13 across languages plus css units_documented. Re-pinned before tagging v2.9.0 so the release batch and the archetype retrain consume corpus-calibrated credit for the new languages. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
The re-pin clears the old sub-1.0 safety-family coefficients (haskell 0.6667, embedded_python 0.5, ...) to exact-on-plant, so safety_score / Error & Exception Exposure shift by design across the crucible corpus (~55 leaves full-precision, ~61 zero-dep). Blessed via tests/tools/update_golden_master.py per mode: full-precision in a [full]-extras env, zero-dependency in a bare env (no numpy -- the first local attempt in the lean v6 venv wrongly baked numpy-present into the zero-dep fixture and was reverted). Both legs of test_golden_crucible pass locally against the new fixtures. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ontract) The zero-dep crucible leg installs PyYAML (documented in golden-crucible.yml since #1104: zero-dependency has only ever meant tiktoken/pandas/xgboost absent). The previous bless ran in a fully bare env, flipping the pyyaml Missing-Dependencies leaf and the api_mapper swagger results (swagger_parse_error vs success + 377 shadow APIs). Re-blessed in a PyYAML-only env; test passes locally in that env. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Regenerates
gitgalaxy/standards/fidelity_table.py(+ the generated wiki tables) from keyword-rosetta main @83fd3487. The committed pin (6377eb59, 2026-09-04, 46 languages × 19 signals) predated the CICS wave — pli, bms, db2_sql, hlasm, rexx (rosetta #135–#138) — and theunits_public/units_documentedsignals.Now 51 languages × 21 signals, 33 cells below 1.0:
branch/state_mutation/io/…) all cleared to exact-on-plant — the alignment campaign's rule fixes, now certified by the table.units_public12/13 across languages, plus cssunits_documented.Generated via the documented flow (tool →
ruff format→ commit);--checkgreen,tests/core_engine/test_language_strictness.py19/19, ruff audit clean.Sequencing: merge before tagging v2.9.0 so the release batch and the archetype retrain consume corpus-calibrated credit for the new languages.
🤖 Generated with Claude Code