release: v1.18.2 — live-arm diagnostic evidence for rate-limit failures; serving still off - #58
Merged
Merged
Conversation
Release prep for the Stage 1 comparator diagnostics already merged in #57: version bump across pyproject.toml, both server.json fields and the generated uv.lock, plus one changelog entry. Also corrects the v1.18.1 heading date from 2026-09-03 to 2026-09-02. That entry was re-dated ahead of publication in #55 on the understanding that 2026-09-03 was the intended date; the release then published on 2026-09-02, leaving the changelog a day ahead of the artifact it describes. The date now matches what actually shipped. No behaviour change: the diagnostics are confined to the out-of-band comparator, which stays outside the request path, and snapshot serving remains off.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Release prep for the Stage 1 comparator diagnostics already merged in #57. Version surfaces plus one changelog entry — no code change.
Changed
pyproject.toml1.18.1→1.18.2server.jsonpackages[0]) →1.18.2uv.lockproperty-shared v1.18.1 -> v1.18.2CHANGELOG.mdChangelog entry
## v1.18.2 (2026-09-02) — live-arm diagnostic evidence for rate-limit failures; serving still offThe comparator now records why a live observation failed, not just that it did: HTTP status and reason, an allow-listed subset of response headers (
Retry-After,RateLimit-*,X-RateLimit-*), and start/finish timestamps on every live observation so realised request spacing can be read off the report. The first Stage 1 run stopped on an HTTP 429 and preserved only the error string, which left the pacing question unanswerable.Comparison semantics unchanged: one live observation per case, same budget guard, no retry, no resume, no automatic backoff. Body never read, no public model widened, comparator stays outside the request path. Snapshot serving remains off.
Also: a v1.18.1 date correction
The v1.18.1 heading is corrected from
2026-09-03back to2026-09-02.PR #55 moved that date forward to 2026-09-03 as the intended publication date; the release then actually published on 2026-09-02T11:17:49Z, leaving the changelog a day ahead of the artifact it describes. This restores agreement with what shipped. Included here deliberately rather than left to drift.
Validation
./scripts/validate.sh— exit 0: lockfile check, pre-commit all-files, 2011 passed, 27 skipped. Focused release-manifest/version tests: 10 passed.Scope
No code, configuration, secret, flag or serving change. Nothing is published or deployed by this PR — publishing v1.18.2 is a separate approval, and only that triggers PyPI plus both Fly deploys.