Setting up a first local model on an NVIDIA GPU installs the container toolkit
A first local model set up with `omarchy agent local use` also installs nvidia-container-toolkit when an NVIDIA GPU is found, so the container reaches that GPU. The change merged into quattro.
GB10 desktops already get the toolkit as a default package. N1x laptops do not, so a model there had no GPU device. David Heinemeier Hansson checked an N1x with everything installed, and the edited wrapper went straight to the engine without asking for sudo. After installing the toolkit, the CDI spec appeared without a reboot.
Version 4.0.6 is the latest release.