ci: keep telemetry failures out of the fetch job status - #107
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Merge as is — freshness lookup failures degrade without failing the fetch jobThis PR should merge as-is because the base workflow runs its GitHub API lookup and date conversion inside an Worth noting, not blocking
Merge state: The latest state reports the PR open, non-draft, and mergeable. Eight of nine checks have completed, with seven successful and one neutral; one check remains in progress. No gating human review or unresolved review thread is reported. Evidence · 1 verified · 3 not verifiedVerified
Not verified
Read the full changed workflow, exact diff, base workflow, sibling Axiom workflow, root contribution rules, package scripts, repository-wide related identifiers, and available PR state and comments. No tests, workflow runs, or PR code were executed. One non-blocking test-coverage conclusion was recorded; no prior review findings or threads were present.
Comment Review by Rupic. |
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
modelpedia-api | fef1d64 | Commit Preview URL Branch Preview URL |
Aug 31 2026, 05:54 AM |
There was a problem hiding this comment.
1 inline comment below, on .github/workflows/fetch-models.yml.
Full review: #107 (comment)
problem
the telemetry step runs under the workflow's fail-fast shell: a GitHub API outage, empty commit list, or date parse failure exits the
if: always()step non-zero and marks the whole fetch job failed, contradicting the step's warning-only degradation design (flagged by review on #103).change
guard the freshness computation: a failed
gh apior unparseable date degrades todataAgeHours: -1("unknown") instead of killing the step.-1can never false-trigger the planned> 48hstaleness monitor.verification
yaml lints clean; the ingest call already degrades to a workflow warning.