Update to HA 2025.12 on dietpi OS

Installation
1

Home assistant on DietPi (v9.20.1, arch: armv7l) (possibly divergent from most users here). Hopefully this helps other with troubles updating to
2025-12 or newer.

Dietpi does not automatically update Home Assistant, but provides a bash script (Home Automation Software Options - DietPi.com Docs) to update from the PyPI repository.

For the update to Home Assistant 2025.12.5 (from 2025.9.4), the procedure described in the link above failed.

  • Tracked down to a failure in building the package tiktoken, due to an outdated version of Cargo.
  • Solved by (re-)installing rust (rustc 1.92.0) by user homeassistant.

This enabled to successfully update Home Assistant via the the Dietpi update script:

sudo -u
homeassistant dash -c ‘. /home/homeassistant/pyenv-activate.sh;
exec pip3 install -U homeassistant
1 Like
2

DietPi user here - love it!

Thanks for sharing.

Is the DietPi script installing the now depreciated HA Core or Supervised versions (I can’t remember)?

I used to use the DietPi HA script but changed to having DietPi install Docker and Docker compose and then running HA Container

1 Like
3

Thank you.
As the DietPi script installs Home Assistant from the PyPI repository, this does in my understanding correspond to the (now depreciated?) HA Core.
@frenck is listed under the maintainers also of the PyPI repo, and its latest version 2026.1.0 was just released Jan 7, 2026.
:point_right: So I assume there would be a deprecation notice well in advance?

:warning: Nothing essential for survival, but with growing numbers of integrations and appliances, I assume I would not be the only one with some troubles …

4
5

Thank you, indeed missed this one :face_with_open_eyes_and_hand_over_mouth:
This announcement states that after release 2025.12, the only supported installation methods are Home Assistant OS or Container.
Yet there is the PyPI repository with the latest version 2026.1.0 just released on Jan 7, 2026.
→ Will this be maintained?

6

As far as I know, HA core will still work in the future, as it will still be used internally for developing HA as I understood. But there is no support for it anymore for end users. So if something doesnt work, that is related to the installation method of using HA core, bug reports will be closed automatically.

Only HA OS and HA containers are officially supported.

I am currently trying to install unsupported HA core with dietpi-software on an outdated and unsupported Rasperry Pi 1B with 512MB of RAM that runs dietpi 10.0.1 with debian 12 bookworm with the new kernel (6.12.62+rpt-rpi-v6 #1 Raspbian 1:6.12.62-1+rpt1~bookworm (2026-01-19) armv6l).

I ran into quite a few issues related to memory that I seem to have fixed for now, after many failed attemps. The install script gave me errors and compiled python 3.13 at least 10 times. So I am trying this for at least a week, as compiling python alone takes maybe around 4 hours or so.

It took me a while to understand from the failed install logs that the problem was either not enough RAM or not enough disk space.

I increased the usable RAM to 475MB by allocating only 16MB for the GPU and configured a swap file of 1.5GB and unmounted the tmpfs on /tmp.

Right now the install script is running for 22 hours and “Building wheel for grpcio (pyproject.toml): still running” for quite a while.

RAM usage is around 100MB to 450MB and swap file around 250MB. The disk access to /tmp seems to slow it down a lot.

I hope this installation finishes at some point, it would be the first time that HA runs on this device under debian 12 bookworm.

Before I was using debian 11 bullseye and dietpi 9.20.2 and I could install HA “succesfully”, but only version 25.1, because only sqlite <3.40 is availabe under debian 11. It was running kind of fine like that, but some integrations could not be installed and even if they worked, they had to be compiled each time for a couple of hours “on demand”. The compiling under debian 11 was faster, as I didnt have to use a swap file there and had tmpfs mounted on /tmp, but also got a lot of errors during compiling due to low memory on /tmp, so some parts were missing but HA did kind of run fine at least.

This obviously isn’t supposed to be a perfectly running HA system, its just supposed to support a few basic features that don’t require as much from the hardware.

It’s not a great idea to do this

I just want to know what’s still possible today with these low specs and this old hardware and if it’s still usable for some use cases.

7

Well. Ha REQUIRES a minimum of a GB ram. Part of the reason they dropped support for the Pi3 and certain Pi4…

Your setup will run almost entirely off swap. For a system that heavily uses a dB…

Read: IF you get it to boot. (which I doubt) you’ll be beating the living hell out of storage which is an order of magnitude slower than ram. Even on SSD before you even run anything…

1 Like
8

I think you’re right! This might take forever to compile.

It’s still compiling, now for 44 hours! It seems to be still working on grpcio.

Also /tmp seems to be running out of disk space, so I will have to restart compiling again after repartitioning the disk.

/tmp is already 2GB big and only 1GB is free at the momemt.

1 Like