- Schema as code. The
.atlfiles in git are authoritative; the database derives from them. - The apply path. The checkpoint, the apply, hot reload, migration ownership, and the boundaries.
- The typed query surface. What the generated per-entity methods can express.
- The generated client. Committed code, stable
proto numbers,
generate --check. - Schema versioning. The append-only version registry: history, blame, staleness.
- Change approval. Classes, tiers, rehearsals, protections, freezes, overrides.
- The console. What the browser view shows, and what deciding requires.
- Custom queries and procedures. SQL beyond the typed surface.
- Caching and invalidation. The two caches and the transactional outbox.
- Tenant isolation. What
partition byenforces, what it trusts, and where it stops. - Jobs and workflows. Typed background work and multi-step orchestration.
- Ephemeral data. Typed, expiring stores with no table behind them.
- The sandbox. Disposable test databases holding schema only, never production data.
Concepts
The concepts pages cover schema as code, the apply path, versioning, change approval, caching, tenant isolation, jobs, and the sandbox.