SQLite

The embedded adapter is the default for local development and for anything self-hosted on Bun or Node with a writable disk. It applies the bundled migrations on construction, so there is no separate setup step.

import { sqlite } from "@heyo-sh/heyo-status/db";export const database = sqlite(process.env.HEYO_STATUS_DB_PATH ?? "status.sqlite");

import { heyoStatus } from "@heyo-sh/heyo-status/config";import { database } from "./heyo-status.database";export const statusPage = heyoStatus({ database, title: "Acme Status" });

Driver selection

sqlite() picks a driver for the runtime it finds:

Both are resolved lazily through createRequire, never a static import, so a bundler cannot hoist bun:sqlite into a build that will run on Node — or into a Workers bundle, where neither exists.

Options

sqlite("status.sqlite");sqlite({ file: "/var/lib/acme/status.sqlite" });sqlite({ file: ":memory:" });sqlite({ client: new Database("status.sqlite"), migrate: false });

:memory: is ideal for tests: each instance starts migrated and empty, and disappears with the process.

Extra methods

Beyond the shared adapter contract, the SQLite adapter exposes:

const results = await database.listResults({  names: ["API"],  from: new Date("2026-01-01"),  to: new Date("2026-02-01"),  limit: 500,});database.close();

listResults() reads raw rows — useful for exports and ad-hoc analysis — and close() releases the connection in a test teardown.

Operational notes

• The file must live on a writable, persistent disk. Serverless filesystems are ephemeral or read-only, which is why neither Vercel nor Cloudflare offers SQLite in the generator matrix.
• Back it up like any SQLite database: copy the file while the process is idle, or use sqlite3 .backup.
• One process should own the file. A single container or VM running the framework server plus the scheduler is the intended shape.
• heyo-status db migrate reports that the embedded migrations are applied automatically; there is nothing further to run.

Where it fits