Documentation
Installation
Clone the repo and run one command. It sets up dependencies, models and the interface for you.
1. Clone
git clone https://github.com/emreyvz/overseer.git
cd overseer
2. One command
The launcher installs the Python toolchain (uv), syncs dependencies, fetches the AI models, builds the interface, and starts the app:
# macOS / Linux
./overseer.sh
# Windows
overseer.cmd
If Node.js is installed you get the desktop app; otherwise Overseer opens in your browser at http://127.0.0.1:8787. The first run downloads model weights into models/; after that it can run fully offline.
Prebuilt installer
If you would rather not clone anything, the latest release carries a one-click installer for Windows (Setup.exe) and for Linux (.AppImage or .deb). It is a thin installer: the AI runtime is fetched on the first launch, matched to your hardware, so allow it a few minutes before the window comes up.
There is no macOS download. Apple's Gatekeeper marks any un-notarized app "damaged and can't be opened", which looks like a corrupt file rather than the policy block it really is, and notarization needs a paid Developer ID. So on a Mac you either run from source with ./overseer.sh above, which is unaffected, or build the installer yourself in one command, where the block never applies.
Manual setup (optional)
Prefer to do it yourself:
uv sync # Python dependencies
cd web && npm install # frontend dependencies
Then run the backend with python main.py (or uv run python -m server) and the desktop shell with cd web && npm run desktop.
For CUDA acceleration, install a PyTorch build matching your CUDA version. Overseer falls back to CPU automatically if no GPU is found.
Continue to Quick Start.