prisma.cursorrules

# Prisma Cursor Rules

You are an expert Prisma developer. Follow these rules:

## Schema Design
- One schema.prisma file. Use prisma-merge for multi-file schemas
- Explicit relation names on ambiguous relations
- @map and @@map for snake_case DB columns with camelCase code
- Use @default(uuid()) or @default(cuid()) — never auto-increment for public IDs
- Enums for fixed value sets. Dont store magic strings

## Relations
- Always define both sides of a relation
- Use @relation with explicit fields and references
- Cascading deletes via onDelete: Cascade only when parent owns children
- Optional relations (?) for truly nullable foreign keys
- Many-to-many with explicit join tables when you need extra fields

## Queries
- select only the fields you need — avoid fetching entire objects
- include for eager loading relations, select for partial fields
- Use findUniqueOrThrow/findFirstOrThrow instead of null checks
- Pagination: cursor-based (cursor + take) for infinite scroll, offset for pages
- Use transactions for multi-step mutations: prisma.$transaction([])

## Migrations
- prisma migrate dev for development, prisma migrate deploy for production
- Never edit generated migration SQL unless absolutely necessary
- Seed data in prisma/seed.ts — run with prisma db seed
- Handle data migrations in separate scripts, not schema migrations

## Performance
- Index frequently queried fields: @@index([field1, field2])
- Use @@unique for compound unique constraints
- Raw queries (prisma.$queryRaw) only for complex SQL Prisma cant express
- Connection pooling: set connection_limit in DATABASE_URL for serverless
- Use Prisma Accelerate or PgBouncer for serverless connection management

## Type Safety
- Generate client after every schema change: prisma generate
- Use Prisma.* types for input validation: Prisma.UserCreateInput
- Prisma middleware for soft deletes, audit logging
- Use extension API for computed fields and query transforms