Anascummis on your computer

The local source kit includes the app, setup scripts and licenses. It requires Bun and Python, plus a first-time download of dependencies and models.

Download local source ZIP

Extract the ZIP, open a terminal in studio, and follow LOCAL-INSTALL.md:

bun install
bun run setup
bun run model:setup
bun run speech:setup
bun run build

Then run bun run model:start and bun run start in separate terminals. Open http://127.0.0.1:3211.

Windows x64 model setup installs llama.cpp automatically. macOS and Linux need a compatible llama.cpp installation and LLAMA_SERVER. Video tools also need FFmpeg/FFprobe. Allow several GB of disk space; device compatibility varies.

Touch the orb to talk.

Double-Space to type. Drop files onto the orb.
First use downloads local speech and model files.