Control plane
Electros LaunchPad is a local control plane: one GUI and CLI on your workstation orchestrate VMs, model runners, and marketplace services through a local daemon.
Pieces
| Piece | Role |
|---|---|
| GUI | Navigation, wizards, shells, catalogs |
elp CLI | Scriptable twin for instance lifecycle |
elpd daemon | Privileged / long-running backend for VMs and host integration |
Runner (llama-server, MLX, …) | Loads model weights and serves inference |
OpenAI gateway (:11434/v1) | Auth (sk-elp- keys) and routing to local or cloud backends |
| llmfit | Catalog search and host-fit scoring |
Trust boundaries
- Inference backends stay on localhost.
- The gateway can listen more widely so VMs (
192.168.67.1) and optional LAN HTTPS clients can connect. - Remote HTTPS clients must trust the LaunchPad CA (see Credentials).
Coexistence with Multipass
Sockets, data dirs, and binary names differ from stock Multipass so both can run on one machine. Settings → Show Multipass instances controls whether Multipass VMs appear in the LaunchPad UI.