LoginSign Up
← Back to all sparks
D

Drizzle ORM

INFRA · APIS
Velocity3.8

TypeScript ORM for SQL databases

orm.drizzle.team ↗

Drizzle's 1.0 RC cycle pairs a performance rebuild with first-class agent tooling

ormcodecsperformanceeffect-v4dialect-parityagent-tooling
◆Current state
Drizzle ORM is deep in its 1.0.0 release-candidate cycle. Two engineering thrusts dominate: a rewritten internals layer (codecs, JIT mappers, Effect v4) that fixes long-standing data-mapping bugs while cutting query latency, and a push to bring every dialect (Postgres, MySQL, SQLite) to parity under that new system. Alongside the ORM, Drizzle Kit is gaining machine-readable output and an explicit AI-agent surface.
◆Where it's heading
The codec system is the spine of this cycle — it unifies how drivers normalize data and unlocks both correctness fixes and speed. After porting it across dialects (rc.3 MySQL, rc.4 SQLite), Drizzle is converging on a stable 1.0. The newer signal is Drizzle Kit going agent-native: JSON output contracts, a programmatic SDK, an MCP server, and bundled Agent Skills aimed at AI coding assistants driving migrations.
◆Prediction
Expect the RC cycle to wind toward a 1.0.0 stable release once remaining dialect parity (notably the SQLite Effect work) lands, with continued investment in the agent-facing Drizzle Kit surface.

◆Recent moves

  1. 4d ago

    v1.0.0-rc.4

    ⚡ SPARK

    rc.4 brings the SQLite rework in line with the codec system, splits SQLite/MySQL into async and Effect variants, adds a wave of @effect/sql and Turso drivers, and — most notably — ships agent-facing Drizzle Kit tooling: JSON output modes, a programmatic SDK, an MCP server, and bundled Agent Skills.

    View source ↗
  2. 1mo ago

    v1.0.0-rc.3

    rc.3 ports the Postgres-side internals work to MySQL: removes legacy RQBv1, unifies query preparation, switches RQBv2 to array mode, and adds streaming fallbacks. Pure dialect-parity progress toward the 1.0 codec architecture.

    View source ↗
  3. 1mo ago

    v1.0.0-rc.2

    rc.2 refines the new codec selectors for custom Postgres columns and adds SQLite migration conflict detection and proper migration-tree merging for branched histories. Incremental hardening of the codec and migration systems introduced in rc.1.

    View source ↗
  4. 2mo ago

    v1.0.0-rc.1

    ⚡ SPARK

    rc.1 is the architectural turning point: opt-in JIT mappers and a new codec system cut benchmarked latency 25-30%, native Effect v4 support lands, and a breaking new casing API moves case control from the db instance to tables/views/schemas. A Netlify database driver ships too.

    View source ↗
  5. 2mo ago

    v1.0.0-beta.22

    beta.22 is a bug-fix roll-up: Windows migration generation, MSSQL float precision, Cloudflare D1 migration, and several drizzle-kit commutativity-check corrections. Stabilization work with no new user-facing capability.

    View source ↗
  6. 2mo ago

    1.0.0-beta.21

    beta.21 fixes a few migration commutativity edge cases — Postgres enum additions are no longer treated as commutative, and enum values added across branches merge correctly. Narrow correctness fixes in the migration engine.

    View source ↗