Backplane reference-run freshness: warning-only + weekly cadence (self-heal CI) - #2827
Conversation
…ekly cadence Follow-up to #2821. That PR stopped reference-run staleness from failing the REQUIRED suite; this addresses the underlying blocker: staleness still hard-failed the dedicated health-78 lane (its CLI exited 1), so the only way to clear the red was a manual reference-run refresh — which nothing schedules, so it recurs weekly. The contract docs establish the intended model: sibling freshness contracts (langsmith-observability-contract.md) make the scheduled path WARNING-ONLY with opt-in enforcement (enforce_block). The reference_run_evidence is a human-reviewed baseline (carries verifier issue/PR + disposition comment), so a 7-day wall-clock hard-fail on it is mis-calibrated. Align with the documented pattern: - validate_backplane_registry.py `main()`: exit non-zero only on BLOCKING (structural) findings. Reference-run staleness is emitted as a `::warning::` annotation (on stderr, so it surfaces even in --json mode) and is non-blocking by default. New `--enforce-freshness` opts into failing on staleness (mirrors langsmith's enforce_block). Report gains blocking_ok/blocking_count/stale_count and per-finding severity. Structural defects still fail (test-gated). - health-78: add a weekly `schedule:` (warning-only cadence surfacing) and a `workflow_dispatch` `enforce_freshness` input for a deliberate gate. With main() now warning-only, the lane goes green-with-warning on an aged baseline instead of stuck-red — no manual intervention needed to keep CI green. Tests: staleness warning-only (exit 0) with stale_count surfaced; --enforce-freshness fails; ::warning:: annotation emitted; structural defects still block (deliberate-break). 27 pass; black/ruff/actionlint clean. NOT in scope (needs a real cross-repo run, won't fabricate provenance): auto- REGENERATING the evidence by re-emitting the Pension-Data reference conformance run and writing back real provenance. Speccing that separately. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
|
Warning Review limit reachedYou’ve reached a temporary PR review limit under our Fair Usage Limits Policy. Next review available in: 43 minutes Your organization has reached its usage spending cap. Adjust your spending cap in the billing tab. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (3)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Workflow source neededPR #2827 needs either a linked GitHub issue or one valid non-issue Workflow Source before PR metadata automation can manage it safely. Please do one of:
Once a valid source is present, this warning will not be reposted. |
GitHub Actions
Inactive
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: dc86ed204e
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| # scheduled path). Reference-run staleness is a ::warning:: here, not a failure, | ||
| # so an aged reference baseline is visible on cadence without wedging CI. Use the | ||
| # manual dispatch (enforce_freshness=true) for a deliberate freshness gate. | ||
| schedule: |
There was a problem hiding this comment.
The workflow now supports a weekly schedule and manual freshness enforcement, but the inspected inventories still describe health-78-backplane-contract.yml as PR/push-only (docs/ci/WORKFLOWS.md:199 and docs/ci/WORKFLOW_SYSTEM.md:737). Operators relying on those canonical docs will not discover the cadence or enforce_freshness gate, so update both inventories and the backplane contract documentation alongside this trigger change.
AGENTS.md reference: AGENTS.md:L60-L65
Useful? React with 👍 / 👎.
Automated Status SummaryHead SHA: cb1e4f9
Coverage Overview
Coverage Trend
Top Coverage Hotspots (lowest coverage)
Low Coverage Files (<50.0%)
Updated automatically; will refresh on subsequent CI/Docker completions. Keepalive checklistScopeNo scope information available Tasks
Acceptance criteria
|
1 participant
Why (the bigger blocker)
#2821 stopped reference-run staleness from failing the required suite. But staleness still hard-failed the dedicated
health-78lane (its CLI exited 1), so the only way to clear that red was a manual reference-run refresh — and nothing schedules one, so it recurs weekly. That's the underlying blocker: freshness didn't self-heal.Design grounding
The contract docs set the intended model: sibling freshness contracts (
docs/contracts/langsmith-observability-contract.md) make the scheduled path warning-only with opt-in enforcement (enforce_block). Andreference_run_evidenceis a human-reviewed baseline (it carries a verifier issue/PR + disposition comment), so a 7-day wall-clock hard-fail on it is mis-calibrated. This aligns the backplane freshness with that documented pattern.Change
validate_backplane_registry.pymain(): exits non-zero only on blocking (structural) findings. Reference-run staleness →::warning::annotation (on stderr, so it shows even in--jsonmode), non-blocking by default. New--enforce-freshnessopts into failing (mirrors langsmithenforce_block). Report gainsblocking_ok/blocking_count/stale_count+ per-findingseverity.health-78: weeklyschedule:(warning-only cadence surfacing) + aworkflow_dispatchenforce_freshnessinput for a deliberate gate. Withmain()warning-only, the lane goes green-with-warning on an aged baseline instead of stuck-red — no manual intervention to keep CI green.Verification
stale_countsurfaced;--enforce-freshnessfails;::warning::emitted; and a deliberate-break proving structural defects (missing sha256) still block. black/ruff/actionlint clean.--enforce-freshnessexits 1.Explicitly out of scope (won't fabricate provenance)
Auto-regenerating the evidence — re-emitting the real Pension-Data reference conformance run and writing back genuine provenance (run URLs, sha256s,
generated_at) — needs an actual cross-repo run and can't be built/verified blind here. That's the true "self-refresh"; I'll spec it separately rather than fake a timestamp. This PR removes the CI pain (no stuck red, no manual clears) and surfaces the freshness signal on cadence.🤖 Generated with Claude Code