LilMikey-CN/prisma-patterns

Schema conventions

• All models use uuid for primary keys via @default(dbgenerated("gen_random_uuid()"))
• Timestamps: always include createdAt and updatedAt
• Soft delete: use status enum, never actually DELETE rows
• Relations: always define both sides explicitly

JSONB for flexible resource metadata

Resources can be anything (skills, physical items, info, connections). Use JSONB for the parts that vary:

model Resource {
  id          String         @id @default(dbgenerated("gen_random_uuid()")) @db.Uuid
  userId      String         @db.Uuid
  type        ResourceType   // HAVE or WANT
  title       String
  description String
  category    String?        // AI-extracted or user-selected
  tags        String[]       // flexible tags
  metadata    Json           @default("{}")  // JSONB: location, duration, quantity, etc.
  embedding   Unsupported("vector(1536)")?   // pgvector
  status      ResourceStatus @default(ACTIVE)
  createdAt   DateTime       @default(now())
  updatedAt   DateTime       @updatedAt

  user        User           @relation(fields: [userId], references: [id])
  // ...
}

The metadata JSONB field holds variable attributes:

{
  "location": "Melbourne, AU",
  "availability": "weekends",
  "estimatedHours": 3,
  "experienceLevel": "intermediate"
}

pgvector setup

Add the extension in a migration:

CREATE EXTENSION IF NOT EXISTS vector;

Use Unsupported("vector(1536)") in Prisma since it doesn't natively support vector types. Query vectors with raw SQL via prisma.$queryRaw.

Migration workflow

1. Edit schema.prisma
2. Run npx prisma validate
3. Run npx prisma migrate dev --name descriptive-name
4. Check generated SQL in prisma/migrations/
5. Never edit migration files after they're applied

Seed data

Always maintain prisma/seed.ts with realistic test data. Run with npx prisma db seed. Include at least 20 users with varied resources for testing match quality.