Repository navigation
What's new
- Apple Intelligence on the Mac. On a Mac with Apple silicon and macOS 26 or later, Rupert can use Apple's built-in model for inbox suggestions and Ask Rupert: nothing to install, nothing leaves the Mac, and answers come in seconds. Choose it in Settings ▸ Local AI ▸ Server; a new install on such a Mac starts with it. A larger model in LM Studio or Ollama still files and answers better.
- Local AI, not just LM Studio. Rupert works with any app that runs AI models on your computer behind an OpenAI-compatible server. It's tested with LM Studio and Ollama; others, such as llama.cpp's server or Jan, should work too.
- Settings ▸ Local AI has presets for LM Studio and Ollama, and Ask Rupert shows "Local AI · model".
- Rupert says where Local AI runs: on this computer, on another computer on your network, or elsewhere. An address outside your network counts as a cloud service: Ask Rupert shows ☁, and inbox suggestions don't use it.
- Thinking models get more room for inbox suggestions, and a clear message if they still run out.
- Fixed: a library folder's … menu ran off the window when the folder's path was long (as Google Drive paths are), hiding Remove from Rupert….
- The screens' framework is updated to SvelteKit 3.
Settings you already have are kept: if you set up LM Studio, Rupert stays on it.
On the iPhone, Ask Rupert comes to the new Ask tab, on iPhones with Apple Intelligence; that update comes through the App Store.
0.4.0 was tagged but never released; this is the first release since 0.3.0.
Installers for macOS (Apple silicon and Intel), Windows and Linux are below. Rupert 0.3.0 on the desktop offers this update by itself once it's published.