Skip to content

Bazzite AI OS Deployment

Run Bazzite Pods with ujust convenience commands on Bazzite AI OS.

Prerequisites

  • Bazzite AI OS installed (Installation Guide)
  • GPU configured (ujust setup-gpu-pods for NVIDIA)

Quick Start

# Show all available commands
ujust --list

# Run ML development pod
ujust apptainer-run-pod-nvidia-python

# Run JupyterLab
ujust apptainer-run-pod-jupyter

# Run DevOps tools
ujust apptainer-run-pod-devops

GPU Setup (NVIDIA)

First-time setup for NVIDIA GPU support:

# One-time setup
ujust setup-gpu-pods

# Verify
nvidia-smi

AMD and Intel GPUs work automatically via /dev/dri.

Pod Commands

Pull Commands

Download pod images (first-time only):

ujust apptainer-pull-pod-base
ujust apptainer-pull-pod-nvidia
ujust apptainer-pull-pod-nvidia-python
ujust apptainer-pull-pod-jupyter
ujust apptainer-pull-pod-devops
ujust apptainer-pull-pod-playwright
ujust apptainer-pull-pod-githubrunner

Run Commands

Start interactive pod sessions:

ujust apptainer-run-pod-base
ujust apptainer-run-pod-nvidia
ujust apptainer-run-pod-nvidia-python
ujust apptainer-run-pod-jupyter
ujust apptainer-run-pod-devops
ujust apptainer-run-pod-playwright
ujust apptainer-run-pod-githubrunner

Workflow Examples

ML Development

# Navigate to project
cd ~/projects/my-ml-project

# Run nvidia-python pod
ujust apptainer-run-pod-nvidia-python

# Inside pod: activate environment and train
pixi shell --manifest-path /opt/pixi/pixi.toml
python train.py

Interactive Notebooks

cd ~/notebooks
ujust apptainer-run-pod-jupyter
# Open http://localhost:8888

DevOps Work

cd ~/infrastructure
ujust apptainer-run-pod-devops

# Inside pod
aws configure
kubectl get pods
tofu plan

Browser Testing

cd ~/test-suite
ujust apptainer-run-pod-playwright
# Connect VNC client to localhost:5900

Workspace Mounting

Your current directory automatically mounts at /workspace:

# On host
cd ~/projects/my-app

# Run pod
ujust apptainer-run-pod-nvidia-python

# Inside pod
pwd  # /home/jovian
ls /workspace  # Your my-app files

Features

  • ujust Commands - Single command to run pods
  • GPU Integration - Automated setup via ujust setup-gpu-pods
  • KDE Plasma - Full desktop environment
  • Atomic Updates - OSTree-based updates with rollback
  • Pre-installed Tools - 450+ development packages

Comparison: ujust vs Docker

Feature ujust (Bazzite AI OS) Docker (Any Platform)
Command ujust apptainer-run-pod-nvidia-python docker run --gpus all ghcr.io/...
GPU setup ujust setup-gpu-pods (one-time) Install NVIDIA Container Toolkit
Pull ujust apptainer-pull-pod-* docker pull ghcr.io/...
Image format SIF (Apptainer) OCI layers
Platform Bazzite AI OS only Any Docker/Podman host

Both approaches run the same pod images - choose based on your environment.

Troubleshooting

Pod Won't Start

# Check if image exists
ls ~/.local/share/apptainer/

# Re-pull the image
ujust apptainer-pull-pod-nvidia-python

GPU Not Detected

# Verify GPU on host
nvidia-smi

# Re-run setup
ujust setup-gpu-pods

# Check GPU compatibility
# See: docs/os/gpu-compatibility.md

Permission Issues

# Ensure you're in correct group
groups  # Should include 'video' and 'render'

# Re-login after group changes

Old Image Version

# Re-pull to get latest
ujust apptainer-pull-pod-nvidia-python

See Also