Description
Newer developer-platform products have GraphQL datasets no comparable exporter covers yet. Record a per-dataset build/skip decision as CFO-0023 did, then build the chosen ones.
Acceptance Criteria
- #1 A per-dataset build/skip decision with reason is recorded
- #2 Chosen datasets are built
- #3 A new dataset or endpoint has a live-verified entry in doc-0003 and an API drift canary probe
- #4 Every new signal and attribute is declared in internal/semconv, listed in docs/signals.md, and has a panel in grafana/build_dashboard.py
Definition of Done
- #1 just check (fmt-check, lint, vet, test, tidy-check, build, vuln)
- #2 just ci before a change that touches the Dockerfile, goreleaser or the image (adds snapshot + image)
- #3 Every new signal or attribute name declared in internal/semconv and listed in docs/signals.md
Implementation Plan
Loop16 chooses aiInferenceAdaptiveGroups for aggregate metrics, not sampled raw-row traffic counts. Root verified schema descriptions and an exact live selection for count, five-minute timestamp and input/output token and inference-time sums. Hyperdrive and Vectorize datasets returned zero rows in seven days and are skipped. No separate raw log collector is chosen. Initial collector stays opt-in and aggregate-only with existing platform series cap, no model or resource identifiers as metric labels.
Implementation Notes
AC1 decided from one settings read and seven-day limit-one existence queries per advertised dataset: both Workers AI sources eligible, Groups chosen for corrected aggregates; raw alternative skipped as redundant for metrics and raw-log expansion not chosen. Eight Hyperdrive/Vectorize sources empty and skipped. Count one is a lower bound, zero is exact. Live schema describes count as total inferences, input/output token sums and total inference time in milliseconds; exact selected fields returned numeric nonnull values. No backend delivery proof or ingestion-lag guarantee is claimed.
Loop16 Workers AI source accepted on exact REV PASS 6783cb01fcf788bedaeac57d8818aab1bb21a312, cherry-picked identically as bb1ba00, pushed d997015505c8b73a97affe9a7bef5c8509d3d312. Public boundary red/candidate, CodeRabbit all14paths and integrated gate passed. Pre-land live contract canary matched; R0 run 37036301862 and landing CI pending. AC3 doc entry/canary conclusion and AC4 dedicated panel remain open. One implementation, zero review-repair/infra. Retention minor justified by existing client gap handling; short configured-window limitation logged separately.
Loop16 live-verified Workers AI field/settings entry added to doc-0003 after exact pre-land canary matched and R0 drift run 37036301862 concluded success at d997015505c8b73a97affe9a7bef5c8509d3d312. AC3 checked. AC4 dedicated Workers AI panel remains open; no new exporter deployment occurred.
Loop16 Workers AI dashboard exact REV PASSb39dc8fa036266131fd953640de758b3f8454418, all four declared counters covered with correct local translation, independent instances/reset/absence semantics. Exact patch pushed65089d58a8068ea4fb7f3f22f80512489a617fc1 and grafana-sync workflow succeeded. Source AC4 checked, taskDone on source, dataset decision and live contract evidence; browser/deployed data and normalized live equality remain separately unverified. No exporter deployment or activation occurred.