@NickM-27 commented on GitHub (Oct 1, 2025):
That's outside the realm of what the genai feature is used for. You're describing periodic single frame analysis. That can be done in an automation…
@vuplea commented on GitHub (Oct 1, 2025):
Hey Nick,
I was searching exactly for a "polling" based LLM analysis - which is affordable with ollama. Use-case in short is detecting animal…
@NickM-27 commented on GitHub (Sep 8, 2025):
I don't think this is something that will really solve the problem. The problem starts with the stream crashing and Frigate reconnecting, which then…
@hawkeye217 commented on GitHub (Sep 8, 2025):
This is already possible in Frigate 0.16 and later using the camera.turn_off service call when using the Frigate HA integration.
@NickM-27 commented on GitHub (Sep 8, 2025):
The review summaries feature, coming in 0.17, uses a set frame count and pulls frames equal distant apart to reach that frame number (as opposed to…
@jmcollin78 commented on GitHub (Sep 8, 2025):
Hello, yes thank you. The documentation is here: https://docs.frigate.video/configuration/live#disabling-cameras
@a-bali commented on GitHub (Sep 8, 2025):
Doing a full restart in such cases always fixed the issue so far. In my case it looks like the stream is missing the video, so it should be easy to…
@jaykin-jmsc-aiml commented on GitHub (Sep 8, 2025):
I'd like to expand on this request with a specific use case that highlights why interval-based frame selection could be valuable, especially…
@bikemike commented on GitHub (Sep 24, 2025):
I agree that something like this would be great. I have similar issues and find that sometimes restarting just go2rtc can correct the issue, other…
@NickM-27 commented on GitHub (Sep 8, 2025):
I see what you're saying, but that is also not really representative of what is possible. For loitering behavior, you should likely be using a…
@jaykin-jmsc-aiml commented on GitHub (Sep 8, 2025):
Thanks for clarifying, that makes sense.
@hawkeye217 commented on GitHub (Sep 8, 2025):
This could get overly complex and not provide much benefit. In testing, vision models are notoriously bad at predicting intent from a series of…