Sound-based music analysis

AudioMuse-AI analyses a music library by how the tracks actually sound, which makes playlists of similar songs possible. It runs as a Docker stack on the TrueNAS server alongside Jellyfin and the music pipeline, and has a web page on its own port.

The GPU I have is one driver too old

My Tesla P4 is on driver 550, but the GPU build of the image needs 580 or newer, so I run the CPU image. It is slower but works.

TrueNAS gives me no sudo on the host. I write the stack files by running commands inside the Dockge container, which has the stacks folder mounted.