NestJS Prerequisite

Before installing the library, confirm that your NestJS project already meets the requirements below.

NestJS application

Runtime

Node 24.x is the primary target: it is what the database matrix runs on. The general CI job also runs 22.x.

Peer dependencies

These packages must be present in your project — they are not bundled with the library. All four ORM/Swagger peers are declared optional, and the root entrypoint loads none of them: import ... from 'nestjs-rest-query' never pulls in an ORM.

@nestjs/swagger is only needed to generate OpenAPI documentation with @ApiDynamicQuery and to try filters from the Swagger UI. Without it, the decorator still registers the endpoint rules; it just emits no OpenAPI parameters. See Swagger.

Databases

Parity is promised over a certified database profile, published in test/profiles/: PostgreSQL 18, MySQL 8.4 LTS and SQL Server 2022, with code-point collation on portable text columns. collectProfileFacts() and checkPortabilityProfile() are exported so you can verify your own database against it, and forRoot({ portability: { enforce: true } }) turns a mismatch into a refusal instead of a surprise.

SQLite is the reference dialect, not a matrix cell: it proves each adapter's compiler implements the plan's semantics. It is what 01-starter-app runs on.

Two column families your schema has to grow

Neither is configuration; both are DDL your application fills in.

• Folded columns. search, ilike and notIlike never emit ILIKE or Prisma's mode: 'insensitive'. They compare a pre-normalised column against a term normalised by the same function, which is what makes the same request return the same rows regardless of server collation. A field used by search with no foldedField makes defineQueryRules refuse to build — the application does not boot.
• Portable order columns. uuid and enum do not order identically across the three database families, so v3 orders by a portableOrderField you declare instead. A uuid primary key needs one for every request, because the pagination tie-break is always applied over the primary key.

Both are covered in detail in Endpoint whitelist rules.

Supported adapters

The adapter is not configured globally. It is decided by the source you pass to execute(), and each source factory lives in its own subpath.

TypeORM

typeormSource(repository) from nestjs-rest-query/typeorm.

• TypeORM configured with at least one database, and entities declared with TypeORM decorators.
• Companion columns follow a convention here, over the entity property path: <field>_folded and <field>_order. The resolver marks any property ending in _folded or _order as internal automatically.
• The model name is derived, not chosen: metadata.name.replace(/Entity$/, '').toLowerCase().
• buildSchemaRegistry(repository) can derive the whole logical registry from entity metadata — but only where a live Repository exists.

Drizzle

drizzleSource({ db, dialect, table, relations }) plus drizzleDatabase({ client, dialect }), both from nestjs-rest-query/drizzle.

• Drizzle 1.0.0-rc.4 or later within the range, on a supported driver.
• The adapter does not read your pgTable. You declare a logical descriptor with createDrizzleTable({ name, model, columns }) and relations by dotted path.
• The dialect is a required argument, because which execution method to call (all() vs execute()) comes from the declared dialect, never from inspecting the object.

Prisma

prismaSource({ client, model, manifest }) from nestjs-rest-query/prisma.

• A generated Prisma Client, passed directly — no cast, no facade of your own.
• A hand-written manifest built with createPrismaManifest, mapping each model to its client delegate and declaring the provider.
• On sqlite and sqlserver, five operators (like, notLike, ilike, notIlike, search) are refused; see the Prisma adapter page.