Hacker News Reader

Comment by timwis | Hacker News Reader

I switched to Home Assistant a couple months ago (because I didn't like the idea of my voice being sent to amazon constantly) and haven't looked back! Soo much more you can do (including immediately using an LLM, if you like, whether in the cloud or local), and so much more control.

View on HN

I am switching out the Alexas in the house to the home assistant voice devices, which can leverage a local LLM without any cloud whatsoever

https://www.home-assistant.io/voice-pe/

The process is still complicated enough to be "enthusiast" (aka nerd) territory but it is getting better with every release. It will still be here in 10 years, nobody can take it away from us.

I think the audience here might be ok with '"enthusiast" (aka nerd) territory' hardware.

How many hours did you spend on this? I want to do it, but I don’t have much time.

For what it's worth, I was worried about that as well, but I found it to be _fun_, and I actually look forward to playing with it, even before bed. Weird, I know, but I'm enjoying it and didn't think I would.

" Why did you pick these default wake words and not something like “computer” or “okay assist”? A wake word should be uncommon in everyday conversations at home or in media, such as music or TV, to minimize the risk of the device activating unintentionally. “Nabu”, “Jarvis”, and “Mycroft” ... "

They hardcoded the wake words in hardware...

Why not just use LLM common sense to say "does this really sound like a purposeful activation?"

Or put a GPU in there, or export the call to your PC like they require for text to speech?

For being a DIY thing, they made it inexplicably hard to D

Because then you'd be running a full speech to text model all the time, and an LLM any time any speech is detected.

Uh yea, I want that to happen. Correct. Why would I want things happening without the model knowing and having context?

Have you seen Star Trek? We should have computer by now. An agent which knows when to tell us things we need to know, not just respond to requests.

>> Because then you'd be running a full speech to text model all the time, and an LLM any time any speech is detected.

> Uh yea, I want that to happen.

Be careful what you wish for.

> Have you seen Star Trek? We should have computer by now. An agent which knows when to tell us things we need to know, not just respond to requests.

Have you read 1984[0]?

Star Trek is fiction. A good fiction IMHO, but a fiction none the less. As such, the writers took "creative liberties" in order to present a storyline for viewers of same.

0 - https://en.wikipedia.org/wiki/Nineteen_Eighty-Four

If you'd read 1984 recently enough you'd know it is based upon the notion of state and social imposed ideology becoming enmeshed, a locally running hub of pure logic is the antithesis of that concept.

A better metaphor would be mephi from yamyatin's We (the original inspiration for 1984) which represents the kernel of u modulated truth which exists outside the calculus of the imposed social matrix

Hilarious that you thought I'd wouldn't be familiar with the work, when you yourself appear to be that person

> If you'd read 1984 recently enough you'd know it is based upon the notion of state and social imposed ideology becoming enmeshed ...

My reference to 1984 was a juxtaposition used to illuminate the fallacy of:

>> Have you seen Star Trek? We should have computer by now. An agent which knows when to tell us things we need to know, not just respond to requests.

Be that as it may, the probability of a dystopian result is far greater than the utopian one quoted.

> A better metaphor would be ...

Since it was not a metaphor, the remaining critique is moot.

They actually have thought it out well. The short of it is you can set up a device to constantly stream to home assistant so that you can use any wake word. That comes with the draw back of more power use and more importantly higher cpu use for each microphone you add. It's still possible (https://github.com/dscripka/openWakeWord)

The alternative for a dedicated low power device is to have trained a model to run on smaller micro controllers so that it can run locally on low powered devices (https://github.com/kahrendt/microWakeWord) this is what they have chosen for their dedicated devices.

This choice also comes with much higher default privacy. Which is great as home assistant offers cloud integration and the fact that they put privacy first in this area makes it much easier to trust they do the right thing in other areas.

A brief run down on the subject https://www.home-assistant.io/voice_control/about_wake_word/

Home Assistant is a great example of the amazing things the open source community is capable of doing. Hopefully they get the color version stuff worked out. It's too bad it appears that most of Amazon's hardware devices can't be turned to the good side and made to work with Home Assistant.

What do you mean color version?

As an aside, I get what you mean in the context of my earlier comment, but just to clarify for anyone reading, you can use alexa and Google home devices with home assistant (eg to send commands and as media players, I believe), but you can't "deamazon"/"degoogle" them and just use the hardware with home assistant, which would be great if we could!

I'd love to de-amazon (neuter) my Alexa's, and somehow flash custom firmware on it. We have some devices around the house, and switching to the linked Home Assistant Voice would be a significant expense now.

You can get an atom echo for $13, if that helps: https://www.home-assistant.io/voice_control/thirteen-usd-voi...

Home Assistant has physical hardware labeled "green" and "yellow".

Beyond that, I assume there's some confusion about what each can and can't do? I certainly can't remember which is which... But I don't actually own one.

Ah, yeah, it was confusing to me when I was trying to get started as well! fortunately, the folks in the hardware channel of their Discord are super helpful.

Here's what I learnt: Green is the new plug and play server. Easiest to setup and will cover the vast majority of what you'll want to do. Yellow is basically green with a couple extra things built in (namely zigbee and thread), plus expansion bays, if you anticipate building on to it, but it requires you to flash it so less user friendly setup but easy if you're a dev.

But most people start out by using an old laptop or PC sitting around. That's what I did, the got a yellow.

Can you create a private network on your router for your spy devices that blocks any traffic to the mothership while allowing something like Spotify? Do they require access to the mothership to work at all?

Data sovereignty in general is always a good decision. I love my HA device, my local inference for LLMs on my 3090, my homelab of services, and the FTP access to my PikaPods workloads.

If you don't have full control over your data, someone else does - and will do as they please with it, eventually.

We mostly use Alexa for playing music (Radio & Spotify) and setting timers. We also use the multi-room music feature where Alexa plays the same audio source on a certain group (great for parties).

Does Home Assistant also have these features?

Are those features worth having a wiretap in your entire home?

No that is why I'm asking. I want a replacement.

Without those features I can just replace Alexa with nothing because we do not use any other features.

It has some of those features. I think I’d attach speakers to the HA Voice you can use them that way but the built-in speaker will not cut it (think gen1 Echo).

They are cool little devices, but they have a while to go before they can replace the Alexa’s.

Ah ok, thanks for the info. Had hoped I could jump on HA Voice right away because I have already a small HA server in my home.

Same here, I was ready to replace 8-10 echo with them in a heartbeat if it was even close to the Echo but it's not quite there yet. I hope v2 is good enough as I'd love to remove the echo's but they are too convenient currently.

Timers works fine.

Music - yes, albeit a bit finicky to set up.

There's an addon called Music Assistant that controls the music, you can connect it to your provider of choice (local files, Spotify, etc), and you can let it connect to several types of sinks.

OOB the HA only knows a few commands (pause, next, etc) but there are some community driven effort to improve the support, such as "Play the album Dark side of the moon". [0].

[0] https://github.com/music-assistant/voice-support

Thanks, good info. I read nothing about the multi-room-music-feature. Do you by chance know if it can be done?

Given the lawsuits against Google from Sonos in the last 5 years, it doesn't seem to be low hanging fruit. Might not exist yet.

Yes, there is functionality to create groups in music assistant. You can even create groups between different types of sinks (so you can combine say Chromecasts together with home assistants media player together with airplay). However, I have not played around with that too much so I don’t know how well the players will sync with each other.

I can’t see any reason why you couldn’t do that. I’ve only been using Home Assistant for a few months, but off the top of my head I reckon I could do that with an automation. The catch is that you may have to set that automation up yourself (either in the UI Or with YAML), and it's less user friendly than an alexa is out of the box. But if you're happy to tinker, as most folks on here are, then the world is your oyster..

This blog post is a great illustration of what I mean: https://blog.jlpouffier.fr/chatgpt-powered-music-search-engi...

Reply on news.ycombinator.com