db

To keep database schema in migrations, establish SQL migration files as the authoritative source for DB structure while ensuring code generation never directly modifies the database layer.

The best way to handle schema evolution is to apply append-only SQL migrations first, then manually update internal ORM types and sqlc queries to match the new schema without regenerating existing migration files.

Proto alignment depends on your project stage: greenfield mode treats proto as authoritative, while migrated mode treats SQL migrations as authoritative, using schema introspection audits to detect divergence between the two.

To safely add a database entity, create a new append-only migration with up and down SQL, apply it to your Postgres database, then update your internal DB entity types and ORM CRUD operations to match the new schema.

To recover from a broken migration in production, write down explicit recovery steps and use a safe roll-forward migration approach, applying a new corrective migration rather than rolling back the existing one.

You should not use generated code to modify database schema when working within a migrations-first contract, as this creates accidental coupling and bypasses the authoritative SQL migration files needed for safe schema evolution.