jmtatsch

No, but I may have found the / my error. In the next step I do 'npm install' and in the dependencies there is a "chromedriver": "latest" which i guess...

I am also troubled by this issue, see https://github.com/evcc-io/evcc/discussions/15329 Losing about 10 % of solar charging share because of this bug.

Today I lost around 2 kwh (20%) while only slow charging (3,5kw) with a charging brick on one phase.

Confirmed working, make sure you alter model_type of vision_config and not one of the other model_types 🙃

I also would like to see audio threshold events and detections in the future.

CI Formatting complaints seem unrelated

Would love to see this. Sounds equal or better than elevenlabs and supports German.

Look at https://github.com/skrashevich/double-take , he has a well maintained fork with support for CodeProject.AI

Which frames are sent to genai processors? The start of recording? The Frame with highest probability of object? Every keyframe? every 10th frame? how can we influence this?

That sounds reasonable. Is there a forced interval relative to the frame rate too? At what resolution are those sent to the genai processor? detection resolution? I just wonder why...