I figured that with 8Gb of RAM, it would take a long time for me to run out of space. But today I am at 85% and it’s creeping up a little more every day.
When I look at the Frigate add-on, I see that the add-on is itself consuming 65% of my RAM.
So, before I hit 100% and everything shuts down, I am wondering about moving Frigate to its own NUC or other micro-PC. My question is, has anyone else done this? Any precautions I need to be aware of?
Frigate has an official docker container which makes it relatively easy to get an instance up and running on any docker-capable system. The processing power of the system is heavily dependent on how many cameras you have and the stream quality (2k, 4k, fps, h264, h265) you need to process and what you want to do with it.
For some sort of frame of reference, I have it running on an i3 10th gen for four 4k cameras. One camera is troublesome and has to be tweaked, so the majority of my processing goes towards ffmpeg having to transcode h265. I have a Coral which I’m using for object detection, though if I didn’t use that, the CPU supports OpenVINO which would cost me some CPU and RAM, but not that much all said (on a system with 16GB and running a bunch of other things too).
I’ve found that an under-powered CPU can cause it to chew through more RAM than it really needs. If you are trying to do some really interesting things with the feeds, having a supported GPU would be where it’s at.
I have a spare Intel NUC and two Coral USB processors where I was planning to install Frigate. My question is in the configuration and integration with Home Assistant. (The few times I used Docker were not pleasant experiences).
regarding frigate integrating with home assistant, it is the same regardless of if it was installed as an addon or on a separate device
When I added a 3rd camera to my Frigate setup I ran into the same issues as you did, @stevemann
I had an old Dell laptop with a broken screen (which I removed) that I just set up with a 2nd HAOS instance and the Frigate Add-On and I’m really happy with that because it means
a) I don’t have to deal with a separate docker setup and all the details
b) I can easily use the Remote Home-Assistant integration
c) I have a ‘playground area’ that allows me to test out some HA stuff that would otherwise interfere with my main HA instance (e.g. comparing Bermuda with ESPresence)
d) I can use the Frigate instance from my Android phone as well (switching with a single 3-finger swipe)
e) I don’t bog down my main instance when I do something that’s a little more taxing for the Frigate machine (e.g exporting clips)
f) improving (or using the more technical term: fiddling with) and restarting/rebooting my main instance doesn’t impact the Frigate one
Hope that helps 
That’s an idea I hadn’t thought of. I already have a Pi3 running an instance of HAOS named “sandbox” for testing purposes, but the NUC would be more capable than the Pi. Still, I also don’t want to learn Docker just for a single package.
How did you integrate Frigate on HAOS#2 so that HAOS#1 automations and cards would connect?
That’s cool. How does this work?
I use the Remote HA as well as the Frigate integration on both HAOS machines - not sure, if that’s absolutely necessary, but I did this such a long time ago, so that I have to honestly say: I don’t remember, but it was easy to set up 
All relevant entities are available on the main HA machine and blueprints work just like if they were using ‘local’ entities.
In the settings of the Android Companion App (not sure about the iOS one) you can add multiple servers and switch between them with a single 3-finger swipe to the right or the left.
GitHub - louislam/dockge: A fancy, easy-to-use and reactive self-hosted docker compose.yaml stack-oriented manager makes it really easy to setup docker containers with an easy to use UI, without all the problems that portainer causes
I have spent the better part of a day trying to get Docker installed on an Intel NUC. I GIVE UP!
I’ll go with the second HAOS + Frigate add-on install next.