Last week, I set-up a navidrome (docker compose) server after tagging my files with MusicBrainz and beets. I serve it over a private network (tailnet) using tailscale serve. It works on all my devices and on iOS with an app called Nautiline. Nautiline has a feature where it will switch between my local network address and my tailnet address seamlessly. It was so simple, I can't actually believe it works. It has CarPlay support and everything. A few clicks and I'm jamming and scrobbling to MusicBrainz. My next goal is to have a local LLM generate smart playlists. Everyone who wants off Spotify, or the other streaming music giants should do this.

> ...scrobbling to MusicBrainz...

Forgive the out of touch questions, but, doing what? And why?

Fabulous! Now just imagine if it supported video too... It would be some kind of... VideoLAN!

How are you feeding the client Lidarr, soularr?

I’ve done the same and included on the same server the equivalent for every type of media: tv shows, movies, ebooks, audiobooks, even YouTube through a sophisticated proxy rotation pool. I have rid myself of every awful enshittified platform and I finally feel free from the bullshit.

Check out audiomuse-ai on GitHub for an open source song vector space embedding system that allows clustering and traversing a huge graph of similar songs, giving you really smart playlists and radios. Smart local LLM playlists are included.

Now I’m on to building a layer to take my song data and use it to query a bunch of apis to allow for “your favorite bands are playing near you” feature that isn’t sponsored.