Database

Andurel v2 uses PostgreSQL through the standalone storage package. One shared database/sql pool backed by pgx serves Bun, sqlc, and River.

Connect through storage

The application constructs *storage.Postgres from a validated storage.Config, then exposes it as the small storage.Connection boundary. Models receive that connection through constructors.

1type Products struct {
2    db storage.Connection
3}
4
5func NewProducts(db storage.Connection) Products {
6    return Products{db: db}
7}

Use db.Executor() for Bun operations and db.DB() for standard SQL consumers. Keep entities, validation, relationships, and persistence methods together in their owning model file.

Schema and migrations

SQL migrations under root migrations/ are the source of truth. The package embeds *.sql files for CLI, tests, and deployment tooling. Bun tags and sqlc output do not replace migrations.

1andurel database migrate new create_products_table
2andurel database migrate up
3andurel generate model Product

Transactions

Transactions bridge Bun and standard SQL. Use storage.RunInTransaction for multi-step workflows. Within the transaction, Executor() serves Bun model work and SQL() can be passed to queries.WithTx for sqlc. This keeps all operations on one transaction and pool.

Factories and seeds

Factories live beside models under models/factories. Named seed compositions live in root seeds/ and are run through cmd/seeds. Check generated factories for drift with andurel generate factories --check --json and sync them explicitly.