AI Models / Arrmes
Arrmes
Use plain language to control your self-hosted media stack.
Arrmes turns requests into structured API calls to Sonarr, Radarr, Lidarr, Bazarr, qBittorrent, Tautulli, Seerr and Bitmagnet. The router runs on your machine and service credentials stay in the application.
What can I ask Arrmes to do?
- Show Sonarr’s queue, Radarr’s history or recent playback activity from Tautulli.
- Find missing episodes, movies, albums or subtitles, or search Bitmagnet’s torrent index.
- Pause or resume a specific qBittorrent transfer, or submit a movie request in Seerr.
Arrmes asks for clarification when a target is ambiguous. Every write requires explicit, single-use confirmation. The capability guide lists supported operations for each service.
Start with a read-only request
Install Arrmes with Python 3.12 or later and follow the authenticated HTTP setup to connect one service. The initial configuration runs without model weights and uses read-only permissions. Try Show Sonarr queue after adding Sonarr.
For learned routing, the default neural router runs on a CPU and its weights occupy 7.52 MiB on disk. Optional GGUF and MLX backends are also documented. The application is MIT licensed. Model weights and adapters carry their own licence terms.
Runtime setup and local routing · Benchmark methods and results