What is seeding?

Seeding is generating test data or fixture data programmatically. Instead of manually creating entities in SQL or database migrations, you write code that generates realistic data — with relations, embedded types, and complex object graphs — ready to use in tests or demo environments.

Seeding vs. other approaches

Approach	Use case	When to reach for it
Seeding	Generate reusable test data with relations	Unit tests, integration tests, CI fixtures, demo databases
Factories	Build one-off entities for a single test	When you need a custom entity per test case
Migrations	Set up schema and seed production data	Schema versions, initial production state
Fixtures	Static snapshots of data	Immutable reference datasets, performance-sensitive tests

Seeding is a factory with relations built in. You write the data shape once on your entities; every test reuses it. Relations are seeded automatically, without boilerplate.

Why decorator-based?

Most seeding libraries require a separate factory file per entity and manual orchestration:

// Traditional: entity + separate factory
const book = new Book()
book.title = faker.lorem.words(4)
book.author = new Author()
book.author.name = faker.person.fullName()
// ...more boilerplate

With @Seed() decorators, the shape lives on the entity itself:

@Entity()
class Book {
  @Seed(() => faker.lorem.words(4))
  @Column()
  title!: string

  @Seed()  // Author is seeded automatically
  @ManyToOne(() => Author)
  author!: Author
}

// One line. Relations, embedded types, circular guards all handled.
const book = await seed(Book).create()

Next steps

• How it works for an overview of the four seeding methods
• Decorating entities to learn the @Seed() decorator syntax
• Seeder suites to organize complex seeding logic into reusable classes