likesync

Mirrors your Navidrome starred songs to ListenBrainz recording loves on a schedule.

One-way: Navidrome stars -> ListenBrainz love feedback. Un-starring a track in Navidrome removes the love on ListenBrainz — but only for loves this tool created (tracked in a local SQLite DB). Loves you made elsewhere are never touched.

How matching works

Each starred song is resolved to a MusicBrainz recording MBID:

1. The musicBrainzId returned by Navidrome (present when your library is tagged).
2. If absent and MBID_LOOKUP_ENABLED=true, the ListenBrainz metadata lookup is queried by artist + title (+ album). Results are cached.
3. Otherwise the track is skipped and logged.

On a large, poorly-tagged library the first run issues one ListenBrainz metadata lookup per uncached track and can hit rate limits — expect a burst of resolve failed logs that clear on later cycles (results, including misses, are cached). To avoid this, tag your library with MusicBrainz IDs (e.g. via Picard) so the musicBrainzId is present, or set MBID_LOOKUP_ENABLED=false.

Configuration

Var	Default	Notes
NAVIDROME_URL	—	required, e.g. https://music.example.com
NAVIDROME_USERNAME	—	required
NAVIDROME_PASSWORD	—	required
LISTENBRAINZ_TOKEN	—	required, from https://listenbrainz.org/settings/
LISTENBRAINZ_URL	https://api.listenbrainz.org	override for self-hosted
SYNC_INTERVAL	1h	Go duration (e.g. 30m, 6h)
RUN_ON_START	true	run a sync immediately on boot
MBID_LOOKUP_ENABLED	true	metadata-lookup fallback toggle
MBID_CACHE_TTL	720h	how long lookup results (incl. misses) are cached; 0 = never expire
STATE_PATH	/data/likesync.db	SQLite file path
LOG_LEVEL	info	debug/info/warn/error
DRY_RUN	false	log intended changes, write nothing

Run

cp .env.example .env   # fill in your values
docker compose up -d --build
docker compose logs -f

Do a DRY_RUN=true pass first to preview what would change.