HailoRT 4.23 & Apps Infrastructure v25.10.0 Released

Hailo-Apps v25.10.0 Released ( Raspberry Pi Trixie OS support)

Hey everyone!

Really excited to share this update with you all – it’s been a long time coming.

We’ve just released HailoRT 4.23 for both Raspberry Pi and Ubuntu, along with a massive new version of the Hailo Applications Infrastructure. I know a lot of you have been waiting for this runtime update, so thanks for hanging in there with us.

What’s new with HailoRT 4.23

The whole software stack is now updated to 4.23 – drivers, runtime, tools, everything.

For Raspberry Pi users: You can grab it all through the unified hailo-all installer, which should make things a lot smoother. This release is built for Debian Trixie,and does not support Bookworm. If you want to keep working with Bookworm, you will have to stay with the previous release. The installation flow is more stable now, and we’ve ironed out a bunch of issues that were cropping up.

How to install/update

For Raspberry Pi:

First, install Hailo-all 4.23:

sudo apt update 
sudo apt full-upgrade
sudo apt install dkms # very important**,** otherwise the driver **won't** be installed correctly 
sudo apt install hailo-all

Then clone and install Hailo-Apps-Infra

git clone https://github.com/hailo-ai/hailo-apps-infra.git
cd hailo-apps-infra
sudo ./scripts/cleanup_installation.sh
sudo ./install.sh

For x86 Ubuntu:
Install Halio packages from Dev Zone.
Clone the repo and use the installer script:

git clone https://github.com/hailo-ai/hailo-apps-infra.git
cd hailo-apps-infra
sudo ./scripts/cleanup_installation.sh
sudo ./install.sh

Apps Infrastructure v25.10.0

This is honestly one of our biggest updates yet. We’ve added some features people have been asking for, plus a ton of stability and workflow improvements under the hood.

New apps:

  • Tiling app (beta)Detects small objects in high-res images by intelligently splitting frames. Great for aerial footage. Includes VisDrone MobileNetSSD. Run it with hailo-tiling.
  • Multisource appProcess multiple video sources at once (cameras, files, RTSP, USB, etc.). You can run up to 3 sources on RPi with per-source callbacks. Use hailo-multisource.
  • REID Multisource (beta)Track people across multiple cameras using face embeddings and re-identification. Uses LanceDB for identity storage, optimized for 15 FPS. Try hailo-reid.

Infrastructure improvements:

  • Completely revamped the installation flow – cleaner error messages, better detection, easier upgrades
  • New and improved C++ post-processing functions
  • Overhauled the face-recognition backend (simplified the logic, removed the UI)
  • Fixed GStreamer pipeline stability and video looping bugs
  • Better memory and buffer management across the board
  • Enhanced logging
  • Smarter resource downloading and version handling

Bug fixes:

  • Hailo8L compatibility issues resolved
  • Pipeline looping bug fixed
  • Version and device detection corrected
  • Installation permission issues sorted
  • Various other stability fixes

Check out the full release notes here: hailo-apps-infra/doc/RELEASE_NOTES.md at main · hailo-ai/hailo-apps-infra · GitHub

Important: Repository changes

Going forward, hailo-apps-infra is now our main development repository and will be renamed to Hailo-apps soon. The hailo-rpi5-examples repo will transition to focus more on community projects and will be opened up for community contributions. We’re excited to see what the community creates!

Thanks

Seriously, your feedback over the past few weeks helped shape this release. It’s great to see this community growing and contributing so much.

Let me know if you run into any issues or have questions. Looking forward to seeing what you all build with this!

1 Like

Hello omria,

Thanks for the latest update, May i know how hailo-apps-infra is different from other hailo repos like Tappas and Application code examples ?

I wanted to know whether hailo has multi stream and multi model pipeline specifically in Python? i see only based on Gstreamer and C++ and python has got only single stream single model inferencing, plz let me know if u have any?

Hello,

I don’t understand. How this will work with Raspberry Pi that has pre-built driver version of 4.20 (built into the kernel) which can not be removed?

Hey @Chuck_Rhodes ,

The DKMS will take care of this!

1 Like

Hey @Vishnuvardhan_M,

Tappas is our GStreamer plugins repo (it used to include apps too, but we split those out).

The hailo-apps-infra is going to become hailo-apps. This is where all our GStreamer pipeline applications live, including multisource support (check it out here: hailo-apps-infra/hailo_apps/hailo_app_python/apps/multisource at main · hailo-ai/hailo-apps-infra · GitHub). You can easily run multiple models together using the service.

We’re also expanding hailo-apps beyond just GStreamer - we’ll be adding more general functionality, with plenty more coming in the next release!

Hi Omri in this post

Gilad said that there will be LPR pipeline in the next release of hailo-apps
but you dont have it here does that mean you wont release it at all ?

Hey @Andrey,

We ran into a bit of an issue compiling for hailo8l, but we’ll release it as a patch by end of week or early next week. Sorry for the inconvenience!

Hello, I am seeing this error while running install. RPI5 with Trixie, Hailo 8

:wrench: Running post-install script…
:magnifying_glass_tilted_left: Checking resources directory permissions…
Traceback (most recent call last):
File “/home/beshan/hailo-apps-infra/venv_hailo_apps/bin/hailo-post-install”, line 3, in
from hailo_apps.hailo_app_python.core.installation.post_install import main
File “/home/beshan/hailo-apps-infra/hailo_apps/hailo_app_python/core/installation/post_install.py”, line 11, in
from hailo_apps.hailo_app_python.core.common.config_utils import load_and_validate_config
File “/home/beshan/hailo-apps-infra/hailo_apps/hailo_app_python/core/common/config_utils.py”, line 10, in
from .defines import (
…<28 lines>…
)
File “/home/beshan/hailo-apps-infra/hailo_apps/hailo_app_python/core/common/defines.py”, line 111, in
TAPPAS_POSTPROC_PATH_DEFAULT = subprocess.check_output(
~~~~~~~~~~~~~~~~~~~~~~~^
[“pkg-config”, “–variable=tappas_postproc_lib_dir”, “hailo-tappas-core”], text=True
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
).strip()
^
File “/usr/lib/python3.13/subprocess.py”, line 472, in check_output
return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
**kwargs).stdout
^^^^^^^^^
File “/usr/lib/python3.13/subprocess.py”, line 577, in run

1 Like

Was able to get it compiled by running the uninstall script in the scripts folder, rebooting and running the above instructions again.

1 Like

Hi i made some pipeline using the some old tutorial and its working pretty good yet I would like to know if you will release the alpr in Apps Infrastructure and i already got the hailo-8 because you discontinued hailo-8L(shame)