CodeScrobble

CodeScrobble is a web application that makes it easy to scrobble your CD or vinyl records to Last.fm by scanning the barcode with your smartphone camera.

Next.js 16 (Pages Router) and React 19, Postgres via Drizzle ORM, and Tailwind CSS. Barcodes are decoded in the browser, by the built-in Barcode Detection API where available and ZXing otherwise. Release data comes from Discogs, scrobbles go to Last.fm.

Development

Try it in Docker

Needs only Docker:

make up       # builds and starts everything, then open http://localhost:3000
make reset    # delete local data and seed again
make down     # stop (keeps the database)

This runs the production image against Postgres and a mock of Last.fm and Discogs (e2e/mock-server.ts). Logging in goes through the mock and is instant; nothing is sent to the real services. The seed data (dev/seed.ts) adds a user with a scrobble history and auto-scrobbles, and the mock knows a dozen made-up records (dev/catalog.ts). Their barcodes are at http://localhost:4011/barcodes, to scan with the camera or to open as /detected/<barcode>.

Run from source

Requirements: Node.js 24 (nvm use), Corepack (corepack enable) and Docker for the database.

yarn install
docker compose up -d postgres   # Postgres on localhost:5432
cp .env.example .env            # add your Discogs and Last.fm API keys
yarn dev                        # http://localhost:3000

Migrations run automatically on startup. Outside production nothing is sent to Last.fm. Scrobbles are only logged.

Environment

Variable	Description
DATABASE_URL	Postgres connection string
SERVER_URL	Public URL; used for the Last.fm callback and secure cookies
DISCOGS_KEY, DISCOGS_SECRET	Discogs application
LASTFM_KEY, LASTFM_SECRET	Last.fm API account
MONGODB_URI	Optional. Imports the CodeScrobble 1.x MongoDB data once on startup (see below)

Database changes

Edit src/server/db/schema.ts, then generate a migration with yarn db:generate and commit the SQL in drizzle/.

Tests

yarn test           # Vitest: server code against in-process PGlite, client units
yarn test:e2e       # Playwright against the production build
yarn lint && yarn typecheck
yarn screenshots    # every screen at phone and desktop size, in test-results/screens

The end-to-end tests run the standalone server (as in Docker) against Postgres (E2E_DATABASE_URL, default: the codescrobble_e2e database of the compose Postgres) and a mock of the Last.fm and Discogs APIs (e2e/mock-server.ts). Chromium's fake camera plays a generated barcode video, so scanning is tested for real.

Deployment

Pushing to staging or production runs .github/workflows/deploy.yml. It runs CI, then builds a multi-arch image (ghcr.io/dpuscher/code-scrobble:<branch> and :<branch>-<sha>) and deploys it with the Dokploy API. It then waits until /health reports the new revision. A successful staging deploy keeps a "Staging to Production" PR open; merging it releases to production.

The GitHub repository needs the DOKPLOY_API_KEY secret. On Dokploy, the apps pull the image and need the environment variables above.

Migrating from CodeScrobble 1.x (MongoDB)

While MONGODB_URI is set, the first start copies users, releases, auto-scrobbles and history from MongoDB into Postgres. It runs in a single transaction and is recorded in the app_meta table, so it runs exactly once. MongoDB is only read. Duplicate user documents (from an old login race) are merged. Login sessions are not migrated: users log in once more. After verifying the import, remove MONGODB_URI, the MongoDB and Redis services, and the mongodb dependency.

License

MIT, see LICENSE.md.