Install the host client

The Ballast client runs open models on your GPU and answers jobs that Ballast sends it. It only ever calls out to Ballast over HTTPS. It never listens for a connection, so there is nothing to forward on your router.

What you need

• Linux on x86_64.
• An NVIDIA GPU with 8 GB of VRAM or more, with the driver installed so that nvidia-smi prints a table.
• Python 3.10 or newer with the venv module. On Debian or Ubuntu that is sudo apt install python3-venv.

Windows and macOS are planned next, and the installer stops with a clear message on both. AMD and Apple GPUs and CPU-only machines are not supported in v1.

Install

curl -fsSL https://ballast.outerscope.xyz/install.sh | sh

If you would rather read the script first, curl -fsSL https://ballast.outerscope.xyz/install.sh | less shows it. It is short.

Before it changes anything, the script prints what it is about to do. In order:

1. It checks for Linux, x86_64, nvidia-smi, 8 GB of VRAM and Python 3.10, and stops with a plain message if any is missing.
2. It makes a private Python environment in ~/.ballast and installs the Ballast client there.
3. If Ollama is not on the machine, it installs it with Ollama's official installer, and may ask for your sudo password. If you already have Ollama it leaves it alone.
4. It links the ballast command into ~/.local/bin and adds a systemd user service so the client starts at login. The service does nothing until you link the machine.

Next, link this machine to your account.