Maintainer analytics dashboard (gate precision, calibration, acceptance rate)

[JSONbored]

Phase 3 — Analysis depth, analytics & developer experience

Context

We already COMPUTE the metrics: gate precision + confusion matrix (src/review/parity.ts computeGateEval), calibration (src/review/ops.ts computeCalibration), reversal rate, cycle-time, slop-band calibration, queue health, and a stats feed (src/review/stats.ts). There is no polished maintainer-facing dashboard, and no acceptance-rate metric.

Goal

A self-host maintainer analytics dashboard that surfaces what we already compute, plus acceptance tracking and export.

Requirements

• Dashboard (self-host; integrate with the existing observability/Grafana stack + stats.ts feed) showing: gate precision/calibration, reversal rate, cycle-time percentiles, slop-band calibration, queue health, findings by category/severity.
• Add a finding acceptance-rate signal (did the contributor act on a finding → merged) tied to inline comments + outcomes.
• CSV/metrics export from the ledger. Bearer-gated; self-host only; public-safe.

Deliverables

• Acceptance signal + dashboard views + export endpoint + tests + docs.

Expected outcome

• Operators see measured (not estimated) accuracy and ROI, with data to tune policy.

Effort

L.

[linear-code]

JSO-2047

[JSONbored]

2026-07-01 audit: canonical maintainer analytics/dashboard tracker — absorbing #795's self-host analytics slice and the dashboard config-generator scope (#1673/#1683). #539 is its backend-done UI-page slice.