Skip to main content

Worker Nodes

A master-worker cluster lets you run AI inference on a dedicated machine while keeping the license and Home Assistant integration on your HA host.

Optional

A worker is only needed if you want to offload inference. If your Home Assistant host can run inference on its own, you don't need any of this, just use the standard HA OS App installation.

When to use it​

  • Your HA host is too lightweight to run inference comfortably.
  • You want to keep the camera close to a separate, dedicated machine instead of cabling it back to the HA host.

Architecture​

RoleResponsibilities
MasterThe HAOS app. Holds the license, talks to Home Assistant, serves the Hub UI, and coordinates the worker.
WorkerA dedicated machine. Reads the RTSP stream from the camera, runs the AI inference, and reports results to the master.

The worker enrolls once via a 6-digit PIN generated on the master. After enrollment it reconnects automatically and authenticates each session, so no PIN is needed again as long as its data volume persists.

Current limits
  • One worker per master.
  • One camera in total. Inference runs either on the worker, or on the HAOS host through the Superhome app. The two do not run a camera at the same time.
The master must be running first

Complete the HA OS App installation before adding the worker. The worker needs network access to the master on your LAN.