AI camera analysis

I use Frigate for person detection, and it would be great to use the Frigate-supplied trigger of binary_sensor.front_doorbell_person and a “camera” of image.front_doorbell_person so that the image that is analyzed is the exact one where a person was detected.

image.front_doorbell.person has entity_picture: /api/image_proxy/image.front_doorbell_person?token=<a token>, but unfortunately, I can’t figure out how to get that or the data from it
into the ai_task.generate_data attachment.

Im getting this error , tried to change some things in yaml but seems to stay, what am i doing wrong ? tried another sensor and camera also.

Hi, I would like to select the snapshot instead of the camera. Is that possible? The camera has a much older picture in memory then the snapshot.

What I mean is the event image in stead to the camera.doorbell