Repository navigation
Release Yoke 1.24.0: local benchmark dependencies and durable failures - #17
Merged
Merged
Conversation
HECer
marked this pull request as ready for review
October 4, 2026 17:48
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Prepare Yoke 1.24.0. The full-repository benchmark now installs its own lockfile-bound local dependencies and stops before model dispatch on setup failure. Safe readiness receipts bind the complete npm invocation without storing prefix credentials. Durable, deduplicated failure observations preserve known causes through execution, verification, completion, integration and recovery, including failed gates followed by successful retries. Independent code/security and measurement-data reviews passed. Release-gate baseline before the documentation-only language correction:
b5db2af6630c07ad0eb51a40c0e9a04ec2185e20. Local complete suite: 1,759 passed, two existing Windows skips, zero failures (1,761 defined cases; 254.866 s through the last test). Lint, build, Canon, docs metadata, npm audit (zero vulnerabilities) and package checks passed. An actual final tarball install validated Canon and exercised the packaged helper offline; the rebuilt runtime matches all measured candidate snapshots. Pre-correction CI run 37210475127 passed Linux/Windows × Node 20/24. Runtime-snapshot CI 37208754601 also passed all four jobs. All concrete independent review findings were resolved; this PR remains a release-preparation draft. Measurements: three alternating version pairs, all six original-test acceptance replays passed; twelve native Codex sessions verifygpt-6.1-sol/mediumand match Yoke usage exactly. Median loop time: 1.23 319.236 s; 1.24 318.147 s. The mixed pairwise direction and uncontrolled provider cache/background load do not establish general speed, token or cost superiority. Three alternating fresh/warm dependency-cache pairs passed. Isolated Windows NEXUS: original 25 unit tests, six browser cases and build passed; populated seven-incident desktop verified. The original project and pre-existing dirty main checkout were preserved. See full measured analysis and release validation. Existing ChatGPT-authenticated Codex CLI was used, with child API-key/endpoint overrides removed; additional paid model APIs were not used. Small TaskQueue fixtures do not replace full-product model studies. Prices, provider cache and unobserved approval/guardian waiting remain unknown. Core dependency setup remains unmanaged, readiness receipts do not attest every installed byte, and legacy missing failure metadata remains unknown. Follow-ups for native model identity and duplicate docs test discovery are recorded. Documentation language correction atddcc7e4ad71618839cd0b5ce5d4d5cb1ba4a3b1d: all maintained German reports, design notes and product discussion notes are now English; analysis filenames and links are updated. Independent translation/numerical reviews, docs metadata and package dry-run passed. The evidence manifest distinguishes original source hashes from published Git blob hashes and explicitly discloses the report translation and existing newline normalization. Original measurement sources and runtime code are unchanged. English publication policy is recorded in AGENTS.md and CONTRIBUTING.md. CI for this documentation-only head passed all four jobs; final release gates below apply to the subsequent changelog-only release head.Final release head:
af7b2efe853c92b434e9c25778a987a1491bdbca(changelog-only readiness correction). Fresh localnpm run prepublishOnlycompleted successfully: lint, build, 186 test files, 1,759 passed, two existing Windows skips, zero failures (1,761 defined cases; 286.96 s total test-run duration), docs metadata and package dry-run. Canon validation passed; npm audit reported zero vulnerabilities. A newly built actual tarball was installed in an isolated project; installed Canon and a real offline packaged-helper setup passed, and packaged helper bytes match source. Rebuilt runtime digest matches every measured candidate snapshot. Final exact-head CI covers Linux/Windows × Node 20/24; all four jobs passed. Independent release integration preserves the dirty main checkout and the original NEXUS project.Publication verified: merged as
274672811bcb156cf4afe5feb34b99396955ebd3, whose Git tree exactly matches the reviewed head. Yoke v1.24.0 is public; tag targets that merge commit. Trusted npm workflow completed successfully and published@hecer/yoke@1.24.0with signed provenance. Public registry metadata now reports 1.24.0; an actual isolated installation from the public registry passed Canon validation and real offline packaged-helper setup. Installed helper bytes match the reviewed source. Registry integrity:sha512-+H2BrGp7ifd5xaBHgfwBBlQTwg1cIjK6oAWaVHRuh5O77rgeDthKXO7RCajVujFm+2J11xbfTPZWBIujY4yhxg==.Post-merge CI passed all four Linux/Windows × Node 20/24 jobs on release merge
274672811bcb156cf4afe5feb34b99396955ebd3.