Steward is a WIP all-in-one tool to index, diff, backup and replicate large FLAC libraries. Though largely meant for my personal use, you might find use of the tool.
It effortlessly allows you to backup and restore your music to S3-compatible storage, with audio content-adressable storage so that changes to metadata doesn't require you to upload or download the large files anew.
Apart from running Steward as a CLI on host or inside of a Docker container, a native macOS UI allows you to use Steward for simpler use cases as well.
Scan through the library and write the entries to a compressed file.
./steward index media/music | gzip > library.json.gzRead an entry from the file.
gzip -d < library.json.gz | head -1NOTE: The digest is of the audio frames inside of the FLAC file, not the file itself. As such it is not affected by changing the metadata of the file, changing its name or location. Only re-coding the audio should have an affect, which makes it great for identifying differences in metadata.
As the index is mostly a flat structure, fully relying on the audio digest to be unique for some operations, duplicates in a library requires some care.
For indexing or uploading a library, one does not need to worry about duplicates. But when diffing and subsequently downloading (parts of) a library, duplicates will make it impossible to know for sure what instance of the duplicate matches a song available locally.
For this reason, Steward will complain about duplicates when diffing indexes and refuse to continue. When downloading, Steward would generate identical file names and refuse to continue. The recommended workaround is to create multiple indexes if required, separating duplicate songs or albums in different folders.
.
├── library 1
│ ├── Uniquez
│ │ └── Unique Songs 10
│ │ └── Track 01.flac
│ └─── Mr. Duplicate
│ └── Best of Duplicate Songs
│ └── Track 01.flac
└── library 2
└── Mr. Duplicate
└── Best of Duplicate Songs
└── Track 01.flac
In this example, don't index the root directory, index library 1 and library 2 separately.
Note also, that when force uploading content, duplicates might get re-uploaded as although the remote audio digest might match all duplicates, the file digest can differ between them, leading to re-uploads.
Musicbrainz' excellent GUI tool to work with annotating music files is called Picard. The actor playing Picard is Stewart and this tool (and to some extent Picard) is a steward of FLAC metadata.
{ "Name": "media/music/Genesis/We Can't dance/Track 01.flac", "ModTime": "2026-01-06T19:20:34.989036+01:00", "Size": 30915750, "Metadata": [ "ALBUM=We Can't Dance", "ARTIST=Genesis", "CDDB=aa10c90c", "DATE=1991", "GENRE=pop", "GENRE=progressive rock", "TITLE=Jesus He Knows Me", "TRACKNUMBER=02", "TRACKTOTAL=12" ], "AudioDigest": "md5:e79c8464b88e88ed2bcc6584fa5bcd43", "PictureDigest": "md5:d41d8cd98f00b204e9800998ecf8427e" }