prisma

Purpose

packages/prisma is the database layer. It owns schema.prisma (the single source of truth for every table), the migration history, the seed scripts, and three generators that turn the schema into typed Prisma client code, Zod schemas, and Kysely types.

Directory layout

prisma/
├── schema.prisma                # 2,877 lines, 100 models, 46 enums
├── migrations/                  # 596 directories of historical migrations
├── zod/                         # Generated Zod schemas (output of zod-prisma-types)
├── client/                      # Re-exports of the generated Prisma client
├── extensions/                  # Custom Prisma client extensions
├── selects/                     # Reusable `select` snippets
├── sql/                         # Raw SQL helpers (grants, PBAC, ...)
├── seed.ts                      # Main seed script
├── seed-pbac-only.ts            # PBAC-only seed
├── auto-migrations.ts           # Auto-runs pending migrations on boot in some envs
├── cleanup-pbac.ts
├── delete-app.ts
├── enum-generator.ts            # Wrapper for the prisma-enum-generator
├── run-enum-generator.js
├── docker-compose.yml           # Local Postgres
├── index.ts                     # Exports the configured PrismaClient
├── is-prisma-available-check.ts
├── package.json
├── zod-utils.ts (34 KB)         # Hand-written Zod helpers extending the generated set
├── zod-utils.test.ts
└── __mocks__/                   # In-memory mock client

Generators

schema.prisma declares four generators. The expected output of each is committed-out and regenerated on demand:

Generator	Output	Purpose
prisma-client	packages/prisma/generated/prisma	The TypeScript Prisma client with types
zod-prisma-types	packages/prisma/zod/	One Zod schema per model
prisma-kysely	packages/kysely/types.ts	Kysely-compatible types for raw SQL workspace
prisma-enum-generator	(custom)	Generates a single enums.ts for runtime enum checks

yarn workspace @calcom/prisma prisma generate

Schema highlights

100 models. The most central:

• User — the universal user
• EventType — bookable offering (with self-relation parentId for managed event types)
• Booking + BookingReference — booking + third-party event IDs
• Attendee — non-organizer participants
• Schedule + Availability — recurring availability templates
• SelectedCalendar — calendars to read free/busy from
• DestinationCalendar — single calendar to write bookings to
• Credential — encrypted third-party tokens (the key JSON column is encrypted with CALENDSO_ENCRYPTION_KEY and never appears in API responses; AGENTS.md)
• App — installed app-store entries
• Webhook + Workflow (workflow runtime is being deprecated)
• Host + HostGroup + HostLocation — team event-type assignments
• Membership + Team + Profile — team / org structures
• HashedLink — private booking links
• Task — backing table for the in-process tasker queue
• OAuthClient + Tokens — Platform tenant credentials

46 enums. Among the more important:

• SchedulingType — ROUND_ROBIN, COLLECTIVE, MANAGED
• PeriodType — UNLIMITED, ROLLING, ROLLING_WINDOW, RANGE
• CreationSource — API_V1, API_V2, WEBAPP
• BookingStatus — ACCEPTED, PENDING, CANCELLED, REJECTED, AWAITING_HOST
• WorkflowMethods, WorkflowTriggerEvents, etc.
• CancellationReasonRequirement — MANDATORY_BOTH, OPTIONAL_BOTH, ...

See primitives for narrative coverage of the most-referenced models.

Migrations

packages/prisma/migrations/ contains 596 migration directories — one of the longest migration histories in any open-source TypeScript codebase. Each migration is a Prisma-generated SQL file plus a migration.toml. The .github/workflows/check-prisma-migrations.yml workflow ensures every PR that changes the schema includes a corresponding migration.

auto-migrations.ts runs pending migrations on boot in some deploy modes; production uses explicit prisma migrate deploy.

Seeds

• seed.ts — the main seed for local dev (creates a couple of users, event types, and apps).
• scripts/seed-app-store.ts — seeds the App table from the registry generated by app-store-cli.
• scripts/seed-booking-audit.ts, seed-huge-event-types.ts, seed-performance-testing.ts — purpose-built seeds for specific scenarios.
• seed-pbac-only.ts and cleanup-pbac.ts — PBAC (permission-based access control) helpers.

yarn db-seed invokes seed.ts via the prisma config block in the root package.json:

"prisma": {
  "schema": "packages/prisma/schema.prisma",
  "seed": "ts-node --transpile-only ./packages/prisma/seed.ts"
}

Conventions

• Use select, not include — every Prisma query in repositories must list its columns explicitly.
• Never expose the Credential.key column in any query that flows to a client.
• Wrap multi-write operations in prisma.$transaction([...]).
• Long-running side effects belong in the tasker (packages/features/tasker), not inline.
• Reusable selects live in packages/prisma/selects/ and should be imported instead of duplicated.

Integration points

• Web app + API v2 both consume the generated client.
• Repositories in packages/features/*/repositories/ are the only layer that should import @calcom/prisma.
• Tests import the in-memory mock from packages/prisma/__mocks__ (powered by prismock).
• Kysely types in packages/kysely are used wherever raw SQL is needed (rare; mostly analytics and PBAC).

Entry points for modification

• Schema change: edit schema.prisma, generate a migration with prisma migrate dev, regenerate client + zod + kysely. Add or update repositories, then services, then handlers.
• Adding a reusable select: drop a file in packages/prisma/selects/ and update the repository.
• Custom client behavior: extensions go under packages/prisma/extensions/ (e.g., logging, soft-delete).

Related pages

• primitives — narrative coverage of the most-referenced models
• features — every domain has a repositories/ directory
• reference → data models