In order to use bknd, you need to prepare access information to your database and potentially install additional dependencies. Connections to the database are managed using Kysely. Therefore, all its dialects are theoretically supported.
Currently supported and tested databases are:
SQLite (embedded): Node.js SQLite, Bun SQLite, LibSQL, SQLocal
SQLite (remote): Turso, Cloudflare D1
Postgres: Vanilla Postgres, Supabase, Neon, Xata
By default, bknd will try to use a SQLite database in-memory. Depending on your runtime, a different SQLite implementation will be used.
When creating an app using App.create() or createApp(), you can pass a connection object in the configuration object.
app.ts
import { createApp } from "bknd";import { sqlite } from "bknd/adapter/sqlite";// a connection is required when creating an app like thisconst app = createApp({ connection: sqlite({ url: ":memory:" }),});
When using an adapter, or using the CLI, bknd will automatically try to use a SQLite implementation depending on the runtime:
app.js
import { serve } from "bknd/adapter/node";serve({ // connection is optional, but recommended connection: { url: "file:data.db" },});
You can also pass a connection instance to the connection property to explictly use a specific connection.
app.js
import { serve } from "bknd/adapter/node";import { sqlite } from "bknd/adapter/sqlite";serve({ connection: sqlite({ url: "file:data.db" }),});
If you're using bknd.config.*, you can specify the connection on the exported object.
bknd.config.ts
import type { BkndConfig } from "bknd";export default { connection: { url: "file:data.db" },} as const satisfies BkndConfig;
Throughout the documentation, it is assumed you use bknd.config.ts to define your connection.
When run with Node.js, a version of 22 (LTS) or higher is required. Please
verify your version by running node -v, and
upgrade if necessary.
The sqlite adapter is automatically resolved based on the runtime.
Runtime
Adapter
In-Memory
File
Remote
Node.js
node:sqlite
✅
✅
❌
Bun
bun:sqlite
✅
✅
❌
Cloudflare Worker/Browser/Edge
libsql
🟠
🟠
✅
The bundled version of the libsql connection only works with remote databases. However, you can pass in a Client from @libsql/client, see LibSQL for more details.
bknd.config.ts
import type { BkndConfig } from "bknd";// no connection is required, bknd will use a SQLite database in-memory// this does not work on edge environments!export default {} as const satisfies BkndConfig;// or explicitly in-memoryexport default { connection: { url: ":memory:" },} as const satisfies BkndConfig;// or explicitly as a fileexport default { connection: { url: "file:<path/to/your/database.db>" },} as const satisfies BkndConfig;
Turso offers a SQLite-fork called LibSQL that runs a server around your SQLite database. The edge-version of the adapter is included in the bundle (remote only):
bknd.config.ts
import { libsql, type BkndConfig } from "bknd";export default { connection: libsql({ url: "libsql://<database>.turso.io", authToken: "<auth-token>", }),} as const satisfies BkndConfig;
If you wish to use LibSQL as file, in-memory or make use of Embedded Replicas, you have to pass in the Client from @libsql/client:
bknd.config.ts
import { libsql, type BkndConfig } from "bknd";import { createClient } from "@libsql/client";const client = createClient({ url: "libsql://<database>.turso.io", authToken: "<auth-token>",});export default { connection: libsql(client),} as const satisfies BkndConfig;
Using the Cloudflare Adapter, you can choose to use a D1 database binding. To do so, you only need to add a D1 database to your wrangler.toml and it'll pick up automatically.
To manually specify which D1 database to take, you can specify it explicitly:
To provide an initial database structure, you can pass initialConfig to the creation of an app. This will only be used if there isn't an existing configuration found in the database given. Here is a quick example:
The initial structure is only respected if the database is empty! If you made
updates, ensure to delete the database first, or perform updates through the
Admin UI.
import { createApp, em, entity, text, number } from "bknd";const schema = em( { posts: entity("posts", { // "id" is automatically added title: text().required(), slug: text().required(), content: text(), views: number(), }), comments: entity("comments", { content: text(), }), // relations and indices are defined separately. // the first argument are the helper functions, the second the entities. }, ({ relation, index }, { posts, comments }) => { relation(comments).manyToOne(posts); // relation as well as index can be chained! index(posts).on(["title"]).on(["slug"], true); },);// to get a type from your schema, use:type Database = (typeof schema)["DB"];// type Database = {// posts: {// id: number;// title: string;// content: string;// views: number;// },// comments: {// id: number;// content: string;// }// }// pass the schema to the appconst app = createApp({ connection: { /* ... */ }, initialConfig: { data: schema.toJSON(), },});
Note that we didn't add relational fields directly to the entity, but instead defined them afterwards. That is because the relations are managed outside the entity scope to have an unified expierence for all kinds of relations (e.g. many-to-many).
Defined relations are currently not part of the produced types for the
structure. We're working on that, but in the meantime, you can define them
manually.
If you have an initial structure created with the prototype functions, you can extend the DB interface with your own schema.
All entity related functions use the types defined in DB from bknd. To get type completion, you can extend that interface with your own schema:
import { em } from "bknd";import { Api } from "bknd/client";const schema = em({ /* ... */});type Database = (typeof schema)["DB"];declare module "bknd" { interface DB extends Database {}}const api = new Api({ /* ... */});const { data: posts } = await api.data.readMany("posts", {});// `posts` is now typed as Database["posts"]
The type completion is available for the API as well as all provided React hooks.