For those who want that the rock5 gets supported by Home Assistant, please vote on the following page:
Rock5 - Home Assistant support
4 Likes
Voted.
BTW, homeassistant supervised works like a charm on armbian - tested by myself
1 Like
Voting will not get support.
Upstreaming device support will get hass support.
Hass only add device support for those devices which are fully supported in upstream kernel.
Good luck.
2 Likes
I have Home Assistant running on Rock5b on Debian in Docker since a couple of days.Still a few issues but running very smooth. However, native support for R5b would be great too, so voted as well.
3 Likes
I try but no luck. maybe you can write how you install Home assistant on Rock5B ?
How To Install Home Assistant on ROCK 5B (Core On Debian)
Pre-installation Steps
Step 1: Download & Write Image:
Download the official Debian system image from: https://github.com/radxa/debos-radxa/releases/download/20230201-0944/rock-5b-debian-bullseye-xfce4-arm64-20230201-1418-gpt.img.xz
Download the flash tool, etcher, from: https://www.balena.io/etcher#download-etcher
Flash Image to ÎźSD card.
Insert ÎźSD Card into the socket on the board.
Power on ROCK 5B by adapter with type C port.
Step 2: Becoming Root:
`$ sudo su -`
Step 3: Updating your System:
`$ apt-get update`
`$ apt-get upgrade -y`
`$ apt-get dist-upgrade -y`
Step 3: Install Dependencyâs:
`$ apt-get install \
apparmor \
jq \
wget \
curl \
udisks2 \
libglib2.0-bin \
network-manager \
dbus \
lsb-release \
systemd-journal-remote \
ca-certificates \
gnupg \
apt-utils -y`
Step 1: Install The Docker Engine:
`$ sudo mkdir -p /etc/apt/keyrings`
`$ curl -fsSL https://download.docker.com/linux/debian/gpg | sudo gpg --dearmor -o /etc/apt/keyrings/docker.gpg`
`$ echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/debian $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null`
Update the apt package index:
`$ sudo apt-get update`
Install Docker Engine, containerd, and Docker Compose:
`$ sudo apt-get install docker-ce docker-ce-cli containerd.io docker-compose-plugin`
Verify that the Docker Engine installation is successful by running the hello-world image:
`$ sudo docker run hello-world`
If Docker is working correctly the following message will be displayed:
Hello from Docker! This message shows that your installation appears to be working correctly.
Step 2: Install Home Assistant:
`$ sudo docker run -d --name homeassistant --privileged --restart=unless-stopped -e TZ=UTC -v /homeassistant/config:/config --network=host ghcr.io/home-assistant/home-assistant:stable`
Find IP-Address:
`$ ip addr`
After it has finished running you should be able to access Home Assistant from any browser:
http://your.ip.address.here:8123
1 Like
Hi Eduard, I followed the steps you wrote as a test because my own install of HA Supervised is running Ok on rock 5B but keeps complaining about Apparmor and Cgroup errors and I cannot solve them.
The steps you wrote are great and install HA very swiftly but that is the âbasicâ version of HA, not HA supervised. For supervised you need Apparmor to run or else HA complains about it even though itâs not breaking the application (for now). Your installation steps also install Apparmor indeed but it doesnât run either, nor can it be started manually as far as I can tell.
The installed Cgroup version in this Debian install appears to be group2 but HA Supervised uses Cgroup1 and also complains about it so both issues I had previously remain the same after your installation. My knowledge of and experience with Linux is rather limited, Iâm afraid. Any suggestions on how to solve the Apparmor and Cgroup issues? Thanks a lot!
Hi. The same problem, iâm working on it to resolve âŚ
2 Likes
you also can build it as a python env
or you just compile all
Have you by any chance played with Assistt and Whisper yet? On the Pi4 you can only use the tiny-int8 model, which does not understand things right in German. I am hoping that the AI engine on the RK3588 and processing power will allow for better models and fast processing.
And how did you install it? I just tried and it is failing when Home Assistant starts
I used the Ubuntu images https://github.com/Joshua-Riek/ubuntu-rockchip/releases
Just installed docker and the homeassistant docker run âŚ
I rate his ubuntu images over all other and shame he doesnât do Debian also, but Ubuntu is a UbianâŚ
Not sure why they use whisper as shows its what they can and can not.
Whisper is SotA for its large models and the tiny quantised version skyrockets in WER.
Its also really hard to fine tune and create custom LMâs such as n-grams and its being used because it was easy to implement but not because its any good for that specific use with the models supplied.
Rock5b can run the base maybe the small model but your starting to get latency then but with HA think you will end up with the tiny model.
| large | medium | small | base | tiny | |
|---|---|---|---|---|---|
| English | 0.15 | 0.17 | 0.17 | 0.20 | 0.23 |
| Italian | 0.16 | 0.17 | 0.22 | 0.33 | 0.46 |
| German | 0.18 | 0.18 | 0.21 | 0.27 | 0.37 |
So yeah German with the tiny model 0.37 WER and yes thats sort of stinky.
I sort of swap between support of enthuasatic hobbyists to sometimes a rage at the snakeoil BS as the implementers are very aware of the reality.
Currently its more âYear of the Potatoâ unless you use the Alexa or Google interface.
Interesting. I wouldnât know why they use it. I guess it is the best free choice out there or is there something better?
I actually tried out installing HA on the board yesterday and set everything up. Actually it is quite fast compared to the Pi4. You can easily use the small model and Assist becomes usable. I wouldnât move the Rock5 yet though, since there is no official support an HA can break easily.
There was some error at the end of the installation, but seems it ran through anyway. Everything works now.
1 Like
Its the usual where the big step up to AI, needs resources of what is being provided by hobbyists and coders on limited budgets.
Its being used because it was easy to implement and that much of what we have had in opensource voice-assitants can be viewed as enthuasasm or just BS of the observational bias of choice.
I have a tendency to think the latter as there have been many conversations about this but still they are used.
I have been hoping that the Voice prospector and investment market may have come to a conclusion the bubble has burst and focus on less branding and move to sharing and collaboration to garner a bigger opensource herd.
Its a shame LLMs have so much opensource published via big names as on the voice side its become quite stagnant.
I would say the wenet framework is likely a better option as they actually openly discuss and have discussed and come up with alternative solutions and state why they did.
https://wenet.org.cn/wenet/lm.html
https://wenet.org.cn/wenet/context.html
Its life and there are some guys out there trying for there dream of being a contract developer and have to be frugal with the real truth and just supply what they can and keep aiming high.
From mic arrays with no processing, and a general lack of audio algs due to the very high level math, to how big data has multiple regional models, its all about resources.
We get what is being supplied, because its what they can and many of these single developers are hoping for lotto event investment, to step up to what is required, but are often fishing with myth and BS.
Both Google & Amazon have created a very deep moat around there products as its self-fulfilling due to the size of the herds that use there products that are collating data.
There still is the Google & Amazon API to Home Assistant but really the Home Assistant version is more of a AIY voice kit as the gap to what many users know of commercial voice control is huge.
It still doesnât stop them trying or over emblishing with sales speak as if they do just get a toe in the market the potential revenue and value is huge.