[silversmith]

My philosophy with HA is not to touch it too much.

I update only when there is some feature I absolutely need to have. And even when there is a new feature, give it a month or two to mature before using. I run a specific docker tag with auto-updates disabled. The instance is not internet accessible, so less care needed for security patches. When there is an exciting new feature, I apply that excitement to the busywork necessary for HA updates, and power through.

Furthermore, I use the built-in UI only for administration purposes. Run a custom react app that talks to HA via websocket/rest APIs. Those are pretty stable in my experience. Same with YAML configured automations.

Sounds like a lot of work, but I feel that it's offset by someone else writing the integrations with 3rd party services. When I want to control my new robot vacuum cleaner from my custom dashboard, all I have to provide is the UI. Someone else has figured out how to talk with the vac APIs, how to authenticate requests, and what to do to avoid getting rate-limited by the 3rd party service. I have to deal with only the HA quirks, not every 3rd party service quirk.

[alex3305]

> My philosophy with HA is not to touch it too much.

I get where you are coming from, but for me updating hasn't given me any issues since about a year or so. Especially if you stick with the .2 or higher releases. Those are really stable.

[jamiepenney]

Yeah I have the same approach - I don't update the second a new release comes out, and check the comments to see what the earlybirds have encountered just in case it's a bad release.

Personally I haven't had any issues for a while now, prior to that it was the Python upgrade and all the Z-Wave changes that caught me out. That's all stable now though so updates don't stress me out much.