seedKnownUsers() inserted demo services with random UUIDs and no
idempotency guard, causing duplicates when the seed ran alongside or
after the main seed (which uses deterministic IDs + ON CONFLICT DO
UPDATE on id). The admin seed API had the same issue.
Now both paths use deterministic UUIDs with ON CONFLICT DO UPDATE on
id, making the upsert idempotent regardless of seed order or repetition.
Same pattern already used for clients (40143c4).
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Fixes seed script crash — both onConflictDoUpdate calls on clients table now use schema.clients.id (PK) as conflict target instead of non-unique schema.clients.email. Email added to set clause for both call sites.
Resolves GRO-298. Unblocks GRO-290, GRO-295, GRO-297.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Before: ~5% probabilistic pending invoices meant UAT couldn't reliably
find billing test data. Shedward was blocked from testing Pay Now flows.
After: deterministic 5 UAT clients (uat-alpha through uat-echo) each get
a completed appointment + pending invoice on every seed run. Client
names and emails documented in Shedward AGENTS.md for direct access.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Convert raw inserts to upserts (ON CONFLICT DO UPDATE) for:
- staff: upsert on email (unique constraint)
- services: upsert on id (deterministic UUID)
- clients: upsert on email (unique constraint)
- pets: upsert on id (deterministic UUID)
This fixes the duplicate key violation when re-running the seed
script against an existing database (e.g., after schema migrations
or test restarts).
Note: appointments, invoices, visit logs still use raw inserts
and would need DELETE-before-insert for full idempotency. Those
tables use deterministic UUIDs so a second seed run without
prior DELETE would still fail. This is scoped to the immediate
staff email constraint violation reported.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Add boolean is_super_user column (default false) to staff table.
Update Staff interface in shared types.
Mark first manager as super user in both seed modes.
Update test fixtures to include isSuperUser field.
Co-authored-by: groombook-ci[bot] <ci@groombook.bot>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Closes GRO-38. Adds POST /api/admin/seed (manager-only, gated by SEED_KNOWN_USERS_ONLY) and separates dev vs prod seeding paths. Reviewed and approved by CTO and QA.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Phase 1 — Seed Hardening:
- Replace all Math.random() calls in seed.ts with a Mulberry32 seeded PRNG
(seed 42) so the same data set is reproduced on every run
- Replace crypto.randomUUID() with a PRNG-based UUID v4 generator
- Add manager (Jordan Lee) and receptionist (Sam Rivera) staff members
to seed — previously all staff were groomers
- New packages/db/src/reset.ts drops all tables/enums and re-runs
migrate + seed; exposed as `pnpm db:reset` at root
- Generate migration 0010_impersonation_sessions.sql for the
impersonation_sessions and impersonation_audit_logs tables that were
already in schema.ts but had no corresponding migration
Phase 2 — Test Factories:
- New packages/db/src/factories.ts with buildStaff, buildClient, buildPet,
buildService, buildAppointment and resetFactoryCounters helpers
- Exported via @groombook/db/factories subpath (package.json + vitest alias)
- impersonation.test.ts updated to use buildStaff instead of hand-rolled
fixture objects
Closes#90 (Phases 1 + 2)
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Creates packages/db/src/seed.ts that generates realistic development data:
- 3 groomers + 3 bathers (staff)
- 10 grooming services
- 500 clients with 1-3 dogs each
- ~2500 appointments across 12 months with varied statuses
- Invoices with line items and tip splits for completed appointments
- Grooming visit logs
Run via: pnpm db:seed (requires DATABASE_URL)
Co-authored-by: Groom Book CEO <ceo@groombook.dev>
Co-authored-by: Paperclip <noreply@paperclip.ing>
Barkley Trimsworth