Totally agree, this would be a game changer if added to the Unifi integration. I just got into Unifi and have some AI cams up, integration with HA works great and is passing all the other data EXCEPT face detection which is all I was hoping for 
as others here have mentioned it would open up so many other possibilities! Prefer not to use web hooks but does anyone know why this is the only data missing in the integration??
I totally agree. Webhooks are a workable fallback for getting face‑recognition events from Protect into Home Assistant, but they only fire after the event is fully completed, which introduces a noticeable delay. When this capability is added directly to the UniFi Protect integration, Home Assistant can read the live WebSocket event stream, giving us the recognition data the moment Protect detects it.
That difference matters a lot for real‑time analysis. With immediate access to the recognition event, automations can decide right away whether the door was opened by someone known or unknown, trigger different security or announcement flows, and even enrich LLM‑based scenarios that depend on fast, structured context. Getting the event at the start instead of the end unlocks much more responsive and intelligent behavior.
1 Like
Considering web socket, I don’t see a meaningful delay with whats available today. Protect person detection to home assistant producing a chime in my home is at most one second.
LLM Vision has a web socket connection in the new version and it causes a lot of problems on my robust HA system.
Putting an LLM in-line would make whatever happens much slower. Face ID just isn’t practical in a home scenario except in very carefully constructed scenarios. I assume it’s used commercial as one tool of several for investigations.