Balances the human review workload: shared reviewer-capacity guidance, dashboard and digest rebalancing cues, per-owner queue context, and typed API parity.
One transparent reviewer-capacity modelA shared derived helper distributes ready outputs across decision-capable teammates using the originating task assignee. It reports assigned and unassigned work, target load, queue spread, watch/overdue counts, most-loaded and lightest reviewers, and a bounded move suggestion without creating a second owner field or changing work automatically.
Capacity roster and owner-load context in the approval queueThe dashboard now shows each eligible reviewer's outstanding load and urgency state, highlights when rebalancing is useful, and names a safe next move. Every output owner pill includes that teammate's total waiting count, while the existing audited task-assignment control remains the only reassignment path.
Daily digest catches workload imbalance before policy breachPreview and production digests use the same capacity calculation as the dashboard. Unassigned reviews point to the lightest eligible queue; uneven owned work suggests a bounded move from the busiest reviewer; balanced teams get a calm confirmation. The preheader calls out when rebalancing is needed.
Workspace-wide capacity API and OpenAPI 3.1 parityGET /api/v1/artifacts now returns a filter-independent approvalCapacity aggregate bounded to the oldest 200 waiting reviews with explicit coverage metadata, plus reviewOwnerCapacity on ready rows. OpenAPI and public docs type the full shape, extending the typed-contract cadence to its 117th unbroken rev. No schema change.
Deterministic production typographyInter, Manrope, and IBM Plex Mono now bundle locally from Fontsource instead of downloading during next/font builds. This preserves the existing brand typography while removing the fonts.gstatic.com availability dependency that caused Vercel's rev 209 deployment to fail on 404 responses.
Start free โ there are no metered credits to run out.