Home Assistant 2023.11

There's AppDaemon/NetDaemon if you want to write automations in python or c#, but very few people use them, because it's massive overkill.

In a way HA is fundamentally an operating system for your home. It does target more technical users than Joe consumer but not by much in the grand scheme of things, especially as they intend to make it more and more 100% web driven, sell "plug in and go hardware", etc.

The plug in and go hardware is interesting because when you think about it a user who is intimidated to flash an SD card and stick it in a Raspberry Pi is going to have quite the shock when an upgrade eventually becomes necessary and they encounter the challenge that is trying to upgrade even a many months old HA install (forget about years).

As they target more and more users of less technical ability and sophistication this is going to only get worse and worse. If you look through this thread even the HN crowd (which leans heavily technical) is saying things that essentially come down to "the last thing I need in my life is more time spent on Github issues and random forums". Many of them end up going back around to "full home luddite" mode because they just can't stand to have to debug a light switch.

I appreciate this all too well. Many people who know me (their nerdy friend) are surprised I use iPhone, Mac, etc. I simply tell them something like "when I spend all day doing the hard tech stuff I have to do I need to be able to turn that part of my brain off and use something dumb". When I'm checked out and burnt out yes please give me the Fisher-Price experience that is the iPhone.

The issue with your upgrade approach is the same thing as all rolling releases - the more time that goes by, the more likely an upgrade is likely to break things even worse.

If you peruse the HA community forums, reddit, Discord, etc you will often see advice like "if you're not going to upgrade every month when you do upgrade do it progressively through each release", which is still and arduous and very time consuming process.

Good to know. I wasn't aware there's a cloud. From what I can tell, Yellow wouldn't be necessary if I have a Hue bridge. I definitely don't need GPIO. Most important to me is usability by others, so what would be the most inconvenient failure modes?

I've been thinking about this a little recently. Raw brain dump incoming...

The problem with intent matching now (say with Home Assistant) is that the transcribed speech more-or-less needs to exactly (character for character) match the defined entities (names of lights, etc) as well as the grammar and structure expected by the intent definitions.

Something as simple as a hyphen in "turn-off" currently breaks the matching. That's an easy one to fix, I'm just providing it as an example.

In terms of addressing this, I've kicked around a few ideas:

With Home Assistant at least we can pull all of the entities and supported intents. We have a variety of options to do all kinds of fuzzy matching with these now known phrases. We'd fix the transcription based on something like nearest neighbor (more or less) and send the clean command to HA. There is also a way to handle this in the prompt to Whisper.

A literal text embedding model with nearest neighbor search seems kind of ridiculous but I've been curious about it. The same could be done at the audio level by essentially making an embedding of the Mel Spectrogram of the actual audio (which we could have available) and searching on that.

There are also approaches with a variety of NLP implementations, language models, etc that could be combined or used separately.

We've also considered some more manual approaches - things like an interface that logs sessions where you can (after the fact) find a session/command that went wrong, and basically map it to what you wanted. While this may seem tedious we get some reports that our speech recognition consistently mis-transcribes a given voice command. Implementing a lookup table to correct for this is simple. The biggest challenge to this is how to deal with the UI/UX aspects.

I don't want to run HA on hardware I already have. I want to put a small box with new hardware on my bookshelf where my Google Home used to live until it was replaced with Alexa, which got removed three weeks ago. If HA is not hardware, why do they have multiple hardware platforms for sale? And what is actually the best value for a dedicated HA device? That is unclear. Why would you need more extensible storage or a variety of RAM possibilities for HA? That is not clear from the first few pages.

I'm sure I can take a spare RPi from my bottom desk drawer and get HA running before lunch tomorrow. I also know that's a waste of a RPi if there's purpose-built hardware (and by that, I mean power profiled, no useless Arduino-layout pins, no SSH configuration, ideally no manual installation at all). If I'm not around and HA stops running because RPis sometimes have weird clock issues or someone closes the power cord in a desk drawer (RIP old laptop) or a service was misconfigured or SD flash storage sometimes goes haywire, are my family members going to be in the dark, literally, while I'm away?

A site that doesn't spoonfeed you the step-by-step, dummy-proof onboarding (a la Hue) usually indicates that the product will require equally esoteric maintenance.

My partner only thinks of Raspberry Pi as a food, and I know of it as a collection of known and yet-to-be-discovered vulnerabilities that runs about 300 mA too much all the time. I have a NUC. It's a royal pain in my ass because it's my primary device and if you don't have a very precise combination of drivers and BIOS settings, it overheats and/or refuses to wake from sleep. (It's actually a replacement; the first overheated one too many times.) I will never, ever run HA on a NUC.

Same deal with Docker: Every time I touch containers, I discover the Who's Who? of certificate errors, "jessie/bionic were last in vogue 5 years ago; no version pin, no run" (yesterday's error), "No connection could be made because the target machine actively refused it", system clock is wrong, etc.

I'm not so cruel to subject my family to opening up a terminal or finding the right manpage anytime the smart home starts acting dumb.

-----

As far as residency mattering... Dollars are on the home page. Only after clicking a few links to buy can I find a variety of countries for buying a Yellow or Green, but the place where I live is not actually the best option, price-wise. Unless someone from mediarath is going to phone me up and walk me through the setup or throw in the perfunctory bag of Goldbären, there's no point in paying extra when the guy a few borders over will ship it for less.

People in Denmark seem to be getting screwed worse, at 126 euros per Green. It's one thing if this was an American or European product, but they're all COO China (and it's still not even breakeven to buy bulk from the Chinese vendor, it would be the U.S. price before tax and customs---ultimately more than the Danes).