The CLI

The blendx command comes from @blendx/cli, which an app adds as a dev dependency; it runs on Bun. Run it from the app's folder (bunx blendx ...), or from anywhere with --cwd <app folder>.

blendx <command> [options]

Commands:
  generate  Write the generated files from schema.dbml and the blends
  review    Write review/<resource>.yaml: what each action does, for a human to check
  migrate   generate: write the next schema migration; up: apply pending schema and data steps

Every command accepts --cwd <dir> (run in that app folder) and -h, --help. Without a command, -v, --version prints the version.

Exit codes: 0 when the command succeeded, 1 when it failed (a --check that found drift included), and 2 for a usage error.

Every command starts from blendx.config.ts (or .js, .mjs) in the app folder, and resolves the paths it names against that folder (Configuration).

blendx generate

Run it after every change to schema.dbml, a blend or src/app.ts. It works in two phases:

schema.dbml becomes src/generated/schema.gen.ts: the Drizzle tables and enums, and models. A schema error stops here, with its line and column (schema.dbml:3:1 expected "," or "]" but found "}").
It imports every blends/*.ts (which import that schema) and the app module, and writes the rest:

File	What it is
`routes.gen.ts`	the routes of every listed action, and `AppType` for the typed client
`client.gen.ts`	every table's action descriptors (route, related writes, and default index sort values), the tables its includes point to, and its primary key, for clients that call actions by name (The React client); it imports nothing, so a web app can load it
`register.gen.ts`	registers the app's type, so hooks and policies see the identity's type
`outbox.gen.ts`	the outbox table, once any hook is a `later` hook, so the next migration creates it; otherwise nothing
`drizzle.config.gen.ts`	the drizzle-kit config that `blendx migrate generate` uses: `schema.gen.ts` and `outbox.gen.ts`
`openapi.json`	the OpenAPI 3.1 document

A file is written only when its content changes. A reply that OpenAPI cannot describe prints a warning: line on stderr; warnings never fail the command (Declaring a reply).

--check writes nothing. It prints a unified diff for each file that is out of date (or says it is missing) and exits 1. Run it in CI.

blendx review

Writes review/<table>.yaml for every blend (Review).

--check writes nothing. It prints a diff for each review file that differs from what the blends say, fails on a review file whose blend is gone, runs every example in review/*.examples.yaml, and exits 1 if anything failed.

blendx migrate

blendx migrate generate [--name <name>] writes the next migration into drizzle/, by comparing schema.gen.ts with the migrations already there. It runs the drizzle-kit version @blendx/cli pins, and refuses to run while schema.gen.ts is out of date, so a migration never comes from a stale schema: run blendx generate first. On a terminal, drizzle-kit may ask whether a column was renamed or replaced.

blendx migrate up applies pending schema migrations first, then loads top-level .ts modules from data-migrations/ (or dataMigrations in the config). Each module default-exports one dataMigration(...) whose ID equals its filename. It prints applied 1 migration from drizzle and one applied data migration <id> line for every newly completed data step. A run with no pending data steps prints data-migrations: no pending data migrations.

A data-step failure names both its ID and the reason, such as data migration 20260926_backfill_order_placed_on failed: orders need a date. The step rolls back and remains pending; fix it with a new, higher-versioned step rather than editing an ID that production may already have recorded.

Migrations use drizzle-kit 1.0's layout: one <timestamp>_<name>/ folder per migration, holding migration.sql and snapshot.json. migrate up refuses the older 0.x layout (meta/_journal.json) and says how to convert it.

A server can also apply migrations itself when it starts, with database.migrate(folder, steps) (Deployment). The CLI is what discovers the directory; a server imports and supplies its own explicit steps array.