WSL2 guide: the installer never sets up the NVIDIA Container Toolkit under WSL, so the AI Assistant silently runs on CPU

[rcronk]

The WSL2 guide's GPU table says "NVIDIA RTX/GTX: Yes. Install the latest Windows NVIDIA driver. That's it." On my install (Windows 11, WSL2 Ubuntu, Path A, RTX 5080, driver current, nvidia-smi working in the shell) that was not enough: the AI Assistant came up CPU-only, and gotcha 5's check showed library=CPU. The reason is in install_nomad.sh's GPU detection, and it is WSL-specific, so I think it belongs in the guide rather than the installer (per #572).

Why the installer misses the GPU under WSL2

install_nomad.sh looks for an NVIDIA card with lspci | grep -i nvidia, then falls back to command -v nvidia-smi:

• lspci lists nothing under WSL2: the GPU is paravirtualized through /dev/dxg, not a PCI device the distro can see.
• nvidia-smi lives in /usr/lib/wsl/lib, which is on the user's PATH but not on root's under sudo (secure_path), so the fallback finds nothing either.

So the script prints No NVIDIA GPU detected. Skipping NVIDIA container toolkit installation., the toolkit is never installed, Docker has no nvidia runtime, and nomad_ollama starts without GPU access. Nothing in the UI says so; only gotcha 5's log check does.

Suggested guide additions (Path A, before step 6)

Either run the installer with the user's PATH so its own fallback works:

sudo env "PATH=$PATH" bash install_nomad.sh

or install the toolkit yourself before running the installer, which is what I did:

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list \
  | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' \
  | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
sudo nvidia-ctk runtime configure --runtime=docker --set-as-default
sudo systemctl restart docker
docker run --rm --gpus all ubuntu nvidia-smi   # should print the card

--set-as-default matters: without it the GPU drops out of nomad_ollama after a reboot (#713).

And in gotcha 5, a line that library=CPU on WSL2 almost always means the toolkit was never installed (the installer cannot detect the GPU there), not a version mismatch.

One smaller note for gotcha 3

The command="mount --make-rshared /" line in wsl.conf only takes effect after wsl --shutdown; the guide says so, but if the installer is run before that first restart, the install fails at nomad_disk_collector after /opt/project-nomad and the credentials have been created. A quick check before step 6 saves the rerun: findmnt -no PROPAGATION / should print shared.

Happy to turn this into a PR against wherever the guide's source lives, if that is easier than editing from this post.