PR2 of Phase B. Wires the alert framework end-to-end:
- alert-rules.ts: 10 rule evaluators implemented as pure async fns over
the existing schema. reservation.no_agreement, interest.stale,
document.signer_overdue, berth.under_offer_stalled, expense.duplicate,
expense.unscanned, interest.high_value_silent, eoi.unsigned_long,
audit.suspicious_login fire against real conditions.
document.expiring_soon stays inert until the documents schema gets an
expires_at column. audit.suspicious_login also stays inert until the
auth layer logs 'login.failed' rows (TODO noted in the rule body).
- alert-engine.ts: runAlertEngine() walks every port × every rule and
calls reconcileAlertsForPort. Errors per (port, rule) are collected
in the summary, not thrown — one bad evaluator can't stop the sweep.
- alerts.service.ts: reconcileAlertsForPort now emits 'alert:created'
socket events on insert and 'alert:resolved' on auto-resolve;
dismissAlert emits 'alert:dismissed'. All scoped to port:{portId}
rooms.
- socket/events.ts: adds the three Server→Client alert event types.
- queue/scheduler.ts: registers 'alerts-evaluate' on the maintenance
queue with cron */5 * * * * (every 5 min, per spec risk register).
- queue/workers/maintenance.ts: dispatches 'alerts-evaluate' to
runAlertEngine; logs sweep summary.
Tests:
- tests/integration/alerts-engine.test.ts (6 cases): seeds reservation
→ fires, runs twice → no dupe, adds agreement → auto-resolves; seeds
stale interest → fires; hot lead silent → critical; engine summary
shape on no-data port. Socket emit module is vi.mocked.
Vitest 681/681 (was 675; +6). tsc clean. Lint clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
PR1 of Phase B per docs/superpowers/specs/2026-04-28-phase-b-insights-alerts-design.md.
Lays the foundation that PRs 2-10 will fill in with behaviour.
Schema (migration 0014):
- alerts table with rule-engine fields (rule_id, severity, link,
entity_type/id, fingerprint, fired/dismissed/acknowledged/resolved
timestamps, jsonb metadata). Partial-unique fingerprint index keeps
one open row per (port, rule, entity); separate indexes power
severity-filtered and time-ordered queries.
- analytics_snapshots (port_id, metric_id) -> jsonb cache + computedAt
for the 15-min recurring refresh.
- expenses: duplicate_of self-FK, dedup_scanned_at, ocr_status/raw/
confidence; partial index on (port, vendor, amount, date) where
duplicate_of IS NULL drives the dedup heuristic.
- audit_logs.search_text: GENERATED ALWAYS tsvector over
action+entity_type+entity_id+user_id, GIN-indexed (drizzle can't
model GENERATED ALWAYS in TS yet, so the migration appends manual
ALTER + the GIN index).
Service skeletons in src/lib/services/:
- alerts.service.ts: fingerprintFor, reconcileAlertsForPort (upsert +
auto-resolve), dismiss, acknowledge, listAlertsForPort.
- alert-rules.ts: RULE_REGISTRY of 10 rule evaluators (currently no-op);
PR2 fills in the bodies.
- analytics.service.ts: readSnapshot/writeSnapshot with 15-min TTL +
no-op compute* stubs for the four chart series; PR3 fills behavior.
- expense-dedup.service.ts: scanForDuplicates + markBestDuplicate
using the partial dedup index. PR8 wires the BullMQ trigger.
- expense-ocr.service.ts: OcrResult/OcrLineItem types + ocrReceipt
stub. PR9 wires Claude Vision (Haiku 4.5 + ephemeral system-prompt
cache).
- audit-search.service.ts: tsvector @@ plainto_tsquery + cursor
pagination on (createdAt, id). PR10 wires the admin UI.
tsc clean, lint clean, vitest 675/675 (one unrelated AES random-output
flake passes solo).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
