Drizzle ORM

Drizzle hits the sweet spot between raw SQL and heavy ORMs. You get full TypeScript inference from your schema, and the queries still read like SQL — no magic, no hidden behavior.

Schema definitions are just TypeScript, migrations are straightforward, and the query builder doesn't try to abstract away the database. Type safety is non-negotiable for me: I do not tolerate any or unchecked unknown values in data flows; unknown input is parsed into typed schema before processing.

Works well with the rest of my stack (Bun, Nuxt, TanStack Query) and doesn't fight against edge deployments.

When I use it

• TypeScript projects with SQL databases (including SQLite), from prototypes to production systems.
• Backends that need both query-builder ergonomics and raw SQL escape hatches.
• Teams that want predictable migrations without heavyweight ORM behavior.

When I avoid it

• Projects without SQL databases (for example Redis-only data layers where Drizzle is not the right abstraction).
• Legacy codebases deeply coupled to another ORM where migration is not currently viable.

One real example

I use Drizzle in both small prototypes and larger products to keep schema and query behavior type-safe from the start, then scale without changing data-access conventions.