# Kima

> Kima is a self-hosted, open-source music server for your own files, adding podcasts, audiobooks and vibe-based matching that maps your library in 3D.

- **Category:** Media Server
- **Licence:** GPL-3.0
- **Latest version:** v1.11.0
- **Open source:** oss
- **Languages:** Typescript
- **Platforms:** Docker, Docker Compose
- **Tags:** Music, Metadata, Spotify, Soulseek, Playlists, Podcast Player, Audiobooks, Music Server, Audio Streaming, Opensubsonic, Lyrics, Lidarr, Subsonic, Podcasts, Scrobbling, Playlist import, Soulseek Network, Musicplayer, Slskd, Audiobook Player, Vibe Matching, Lyrics Fetcher, Music Sharing

## Description

Kima is a DIY music server: you hand it a folder of songs and get browser playback on desktop and phone, plus podcasts and audiobooks. Every file Kima reads in the common lossless and lossy formats is identified via MusicBrainz tags, album tracklists and AcoustID fingerprints, and lyrics come from the tags or LRCLIB. Its signature piece is the vibe engine, which rates each track on three axes — a CLAP sound embedding, audio-derived feel (valence and arousal, calibrated against your whole collection) and the artist's MusicBrainz genres. That single similarity score powers recommendations, Song Path transitions, Blend playlists, mood mixes and a 3D map of the library. Music acquisition is first class: album requests go to Lidarr, Soulseek search and downloads work built-in or through slskd, weekly discovery playlists pull music you do not own, and playlists can be imported from Spotify, Deezer, YouTube, SoundCloud, Bandcamp, Mixcloud or an M3U file. Audiobooks stream from a linked Audiobookshelf server with positions synced back, plays can scrobble to your own ListenBrainz account, and ordinary Subsonic clients connect through the OpenSubsonic API with revocable per-device tokens. Several accounts are supported, the first one being admin, with optional 2FA, public playlist share links and listening statistics. Operationally it is a single container bundling the web app, API, Postgres, Redis and both analyzers behind port 3030, with everything else in /data; optional NVIDIA GPU switches speed up enrichment. There is no telemetry, the iOS app is only promised, and the project is young (its first releases date from late 2025) but ships stable 1.x updates regularly.

## Links

- [Homepage](https://github.com/chevron7locked/kima-hub)
- [Docker](chevron7locked/kima)
- [Install and update](https://github.com/Chevron7Locked/kima-hub/blob/main/docs/install.md)
- [Configuration](https://github.com/Chevron7Locked/kima-hub/blob/main/docs/configuration.md)
- [Integrations](https://github.com/Chevron7Locked/kima-hub/blob/main/docs/integrations.md)
- [Subsonic clients](https://github.com/Chevron7Locked/kima-hub/blob/main/docs/subsonic.md)
- [NVIDIA GPU](https://github.com/Chevron7Locked/kima-hub/blob/main/docs/gpu.md)
- [Administration](https://github.com/Chevron7Locked/kima-hub/blob/main/docs/administration.md)
- [GitHub Issues](https://github.com/Chevron7Locked/kima-hub/issues)

---
Source: https://selfhost.directory/project/kima-hub