Feedback/friction telemetry — centralized instrumented capture contract (root fix)
Root fix for feedback/friction telemetry centralized instrumented capture contract. Scope (phased commits OK, not done until all land): 1) falsifiable delivery baseline (activity-vs-feedback, is capture broken); 2) centralized instrumented feedback primitive - no error/warning/guide/nudge/dialog/empty-state/success feedback displayable without it, remove manual setError+logger pairing; 3) canonical queryable taxonomy under category=toast umbrella (error/warning/guide-nudge/dialog/empty-state/success + confirm/cancel where decision-friction matters), PII-free, sane impression dedupe; 4) delivery observability - sendBeacon=false fallback, accepted/rejected/rate-limited/failed health, retry/fallback where safe, synthetic heartbeat so quiet != broken; 5) census + migrate ALL feedback surfaces incl server-rendered banners/dialogs; 6) per-surface coverage lint CI-enforced, no broad allowlists; 7) admin visibility - capture freshness, delivery failures, activity-vs-feedback silence, uncapped trends (not vFriction7d LIMIT 100); 8) prod end-to-end verify incl proving an intentionally failed transport is visible. Elazar-directed, relayed via audit-venus-ca. coder-venus-cc owns design/taxonomy; only genuine product forks escalate to Elazar.
Questions
Activity
-
Stage landed: P2b transport/idempotency pipeline + 10 bounded per-domain codes, SHA 969af25 v0.95.0 live, deploy-verified (PM independent Vercel MCP check). Server-side edge code-typing (ActionError/classify/guardOr) + 196-msg catalog + 45-callsite producer rewire deferred to later phased commits per real collision with venus's existing result-code discriminant (not a coherence/privacy regression — P2b already kills free-text logging). Remaining WI scope items 5-8 (census+migrate all surfaces, coverage lint, admin visibility, prod e2e verify) still open.
-
Lane 2 Scope A landed: domain feedback codes on 4 reachable error toasts (practica.error x3, support.error x1), client-only, no hook/core/schema/server/auth touch. SHA 9bded20/v0.95.2, PM-verified independently (live /api/health 0.95.2 dbOk:true, dpl_GobHTPaYTx2FQRBGwcWXSTBF5h2c, 0 runtime errors 1h). Audit PASS on pre-impl design + pre-push diff + PTD (coder-supplied fallback, audit OAuth-down). Scope B (granular per-msg codes + indexed column, sink migration) parked as separate future WI, loop solo-venus first. WI stays inProgress -- scope items 5-8 + feedbackCode/edge-typing still open.
-
Root fix landed: P2b transport/idempotency pipeline (SHA 969af25), coverage-lint + admin-visibility panel + domain codes on all 4 reachable error toasts (SHA 9bded20/cfcaca6), item-8 prod e2e failure-path proof (canary probe, operator-run: Elazar authenticated, proxy-venus-cc-w drove mechanics, zero clinical-data touch, audit PASS). Scope B (granular per-msg codes + indexed feedbackCode column + full callsite rewire) carved out to VENUS-216, parked, loops solo-venus first.