trailSailServer turns this Mac into the model server for TrailSail.

It installs and manages Splash (native MLX), the llama.cpp router and, when Docker is running, SearXNG for web search. It generates the API keys, keeps the engines running across reboots, publishes them on your Tailscale network only, and runs a small control service so TrailSail on your Mac, phone or tablet can start, stop and switch models without screen sharing.

Before you continue, make sure that on this Mac:

  •  Tailscale is installed and signed in to your tailnet
  •  HTTPS is turned on for your Tailscale network: at login.tailscale.com/admin/dns, check that MagicDNS is on and press Enable HTTPS under HTTPS Certificates
  •  Homebrew is installed (brew.sh)
  •  OrbStack or Docker Desktop is running, if you want web search

You will be asked for your password once. It is used to set how much of this Mac's memory models may use, and to install the small helper that lets you change that later from TrailSail, and for nothing else. The install takes a few minutes if Splash and llama.cpp are not yet installed; the first model download happens afterwards in the background.

When it finishes, a dialog shows a one-time pairing code (also copied to the clipboard). Enter it in TrailSail under Settings → Server → Pair with a server.
