So, I got a few Amazon A-words a few years ago and some smart lights and all that and they were the bees knees... For a week, then the novelty wore off. Two years later, I'm just stuck in my routine of "get ready for bed" and the lamp turns on so I can shut off the room light, then "night light" and it turns red at 3% brightness and shuts off after 20 minutes.

Well, the a-word is just so filled with ads and stuff now. I guess I trust them enough to "not listen to everything I say" but, it's just getting annoying. Also, when my a-word and the lightbulbs are on the same WiFi, I can't turn my light on when the internet is down!

I have an ESX box at home - it's just an old Dell desktop so no crazy resources - but it could run another VM just fine.

1. What are the options for running a server that can listen to me and do things? Obviously I'd need a mic, etc. (I know Windows has more overhead but I'm most comfortable with that, but I know enough Linux to make it work, and I'm good at following documentation).
2. My smart lights are Kasa - so... Would a self-hosted server be able to control those? Would it still need their app? So, would it still not work when the internet is down?
3. What aren't I thinking of I might need to know?