This edition
Beyond sqlc: teaching AI to generate repositories and integration tests in Go
October 07 2025 - 09:30 (45 min )
Using sqlc (a SQL compiler) in Go is great for generating type-safe database access. However, using the structs generated by sqlc directly in the business logic layer would tightly couple the domain to the database schema. To keep things flexible, a repository layer is needed to map between generated structs and domain models. But writing this mapping by hand quickly becomes repetitive.
The talk begins with a review of the evolution of working with Postgres in Go: from the standard library’s database/sql, through Squirrel + sqlx, pgx driver and API, to modern SQL-first approaches with sqlc. It then introduces an AI-powered generator that creates repository layers based on provided interfaces and domain models, leveraging sqlc-generated queries. The generator, playfully named sqlc++, is powered by Claude Code with a custom command and uses a GitHub project with a reference implementation as an example. This reference-based approach reduces LLM hallucinations and ensures consistent code style.
This approach makes it easier to bootstrap a new repository: in addition to what sqlc already requires (SQL schema and queries), one has to define domain models and a repository interface. From there, the sqlc++ generator produces the repository implementation and integration tests, ensuring consistency and avoiding repetitive boilerplate.
LEVEL: Intermediate
The talk begins with a review of the evolution of working with Postgres in Go: from the standard library’s database/sql, through Squirrel + sqlx, pgx driver and API, to modern SQL-first approaches with sqlc. It then introduces an AI-powered generator that creates repository layers based on provided interfaces and domain models, leveraging sqlc-generated queries. The generator, playfully named sqlc++, is powered by Claude Code with a custom command and uses a GitHub project with a reference implementation as an example. This reference-based approach reduces LLM hallucinations and ensures consistent code style.
This approach makes it easier to bootstrap a new repository: in addition to what sqlc already requires (SQL schema and queries), one has to define domain models and a repository interface. From there, the sqlc++ generator produces the repository implementation and integration tests, ensuring consistency and avoiding repetitive boilerplate.
LEVEL: Intermediate
