Right AI Detection modus for only AI trigger motion

[titro]

Hey,

what is the right modus if i 'm only using the trigger motion api event?

Activating the AI detection modus?

And what modus? doesn't matter in that case?

[matteius]

For trigger-only use, the setup is:

• AI detection / detection-based recording: enabled
• Detection model: leave it empty

The detection thread then runs and responds to your /api/motion/trigger calls, but performs no local inference — so nothing is spent on a model whose output you aren't using. Choosing a model means LightNVR also runs that detector on the stream continuously, which is only what you want if you intend to combine local detection with your external events. So it does matter: an unnecessary model here is pure CPU cost.

As of 0.36.5 there's a second option. If you record 24/7 and only want the API events marked on your timeline, leave detection-based recording off entirely. That combination used to be rejected with a 409; now the event is recorded against your existing recording and the response tells you recording_triggered: false.

Goal	Detection-based recording	Detection model
The API event starts and stops the recording	enabled	leave empty
24/7 recording, API events only mark the timeline	disabled	n/a

Both are documented now: docs/API.md → Trigger Motion Event

Related: your #466 also landed in 0.36.5, so you can send the object class and tags along with the trigger and have them show up on the timeline.