On this page
Installation
Install BOSS Offline AI on macOS, Linux, Windows, or with Docker. Offline by default.
Requirements
| Platform | Minimum | Notes |
|---|---|---|
| Linux | Python 3.10 | CPU or NVIDIA GPU |
| macOS | Python 3.10 | Apple silicon or Intel |
| Windows | Python 3.10 | WSL2 recommended |
| Docker | Docker 24+ | Any host |
pip (all platforms)
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install boss-offline-ai
Docker
docker run --rm -it \
-v "$HOME/.boss:/root/.boss" \
ghcr.io/vectorlogic/boss-offline-ai:latest
The volume keeps the downloaded weights so you do not re-download them on every run.
Verifying the install
boss --version
boss doctor
boss doctor checks your Python version, available memory, and whether a
usable accelerator is present. It never contacts the network.
Air-gapped setup
If the target machine has no internet access at all, download the model
weights on a connected machine, copy them to ~/.boss/models/, and run
boss doctor --offline to confirm everything resolves locally.
Troubleshooting
boss: command not found— the virtual environment is not active.- Out of memory — use
--quantize int8to halve the memory footprint. - Slow first run — weights are being written to the local cache; later runs are fast.