Before you start¶

Everything below runs at home on your own laptop. Doing it in advance means you spend the workshop learning instead of downloading.

1. Install Pixi¶

curl -fsSL https://pixi.sh/install.sh | bash

powershell -ExecutionPolicy ByPass -c "irm -useb https://pixi.sh/install.ps1 | iex"

Open a new terminal and check the version. You need 0.77 or newer:

pixi --version

2. Get the workshop material¶

git clone https://github.com/prefix-dev/roscon-2026-declarative-ros-workspaces-with-pixi.git
cd roscon-2026-declarative-ros-workspaces-with-pixi

3. Warm up your cache¶

pixi install --all --manifest-path solutions/01-ros-workspace/pixi.toml

--all fetches both the Lyrical and the Kilted environments, including PyTorch for your machine. It does not download the packages for every other platform.

4. Check that it works¶

pixi run --manifest-path solutions/01-ros-workspace/pixi.toml sim

A window with a turtle should appear. Close it, then check the second distro too:

pixi run --manifest-path solutions/01-ros-workspace/pixi.toml -e kilted sim

If both worked, you are ready.

What you do not need¶

• No ROS installation. No apt, no rosdep, no /opt/ros. If you already have one it will not be used. Do check your .bashrc for a source /opt/ros/<distro>/setup.bash line and turn it off for the workshop: it sets ROS variables in every terminal, and those can shadow the Pixi environment.
• No sudo or virtual machine for the local ROS exercises.
• No GPU for the local exercises through the cross-platform solve. Running the CUDA computation in Exercise 1.9 does need an NVIDIA GPU and a compatible driver; an Apple GPU cannot run CUDA. Use your own compatible Linux GPU machine or join the Brev setup with the instructors to try it during the workshop.

Bringing your own project¶

You are welcome to work on your own ROS workspace instead of the example. Have it checked out and building the way you build it today, and bring your package.xml files. See Migrating from rosdep & colcon for the path we will follow.