Metric queries were written against CLI v53.4.0. Current CLIs (seen on 62.5.0 and 65.2.0) changed three things, so 16 of 32 queries failed with invalid_request, including the Observability Plus canary in collect-signals. Audits then reported Observability Plus as unusable (blocker=all_failed_other) on teams that have it.
- *.count metrics allow only count/unique; the skill used sum.
- Dimensions are camelCase (originHostname, cacheResult, botCategory, requestPath, sourceImageHostname, wafAction, ...).
- Metrics were renamed: speed_insights_metric.{lcp,inp,cls,ttfb_ms} -> speed_insights.{lcp_ms,inp_ms,cls,ttfb_ms}; speed_insights_metric.count was removed (sample volume now comes from the count of LCP samples); middleware_invocation.duration_ms -> request_duration_ms.
queryMetric now retries a rejected query in the current form (modernizeMetricQuery) and maps the response back to the legacy metric, aggregation, and dimension names (restoreLegacyMetricShape). Callers, gates, and the v53.4.0 fixtures are unchanged, and older CLIs keep the original query.
Verified against vercel metrics schema <metric> on CLI 65.2.0. With the
first part of this fix, the canary passed and request-count queries returned
56 rows on a team with Observability Plus.
Repro: vercel metrics vercel.request.count --format json -a sum --since 14d --limit 1 -> {"error":{"code":"invalid_request"}} vercel metrics vercel.request.count --format json --since 14d --limit 1 -> summary [{ vercel_request_count_count: 257 }]
Activity
4This pull request is a draft
Merging is unavailable until it is marked ready for review.
The latest updates on your projects. Learn more about Vercel for GitHub.