Installation
Recommended: One-command setup
Section titled “Recommended: One-command setup”The fastest way to get started is with uv, the fast Python package manager:
# Install uv if you don't have itcurl -LsSf https://astral.sh/uv/install.sh | sh
# Install sparkrun and set up your cluster in one stepuvx sparkrun setupThis runs the setup wizard, which walks you through everything interactively — cluster creation, SSH mesh, CX7 networking, and more. Each phase is optional and can be revisited later.
See the Setup Wizard Walkthrough for a detailed step-by-step guide.
Alternative: Install only
Section titled “Alternative: Install only”If you just want to install sparkrun without running the full wizard:
uvx sparkrun setup installThis creates a managed virtual environment, installs sparkrun, creates a shell-ready command, and sets up tab completion — all in one. You can run the wizard later with sparkrun setup wizard.
Verify installation
Section titled “Verify installation”sparkrun --versionUpdating
Section titled “Updating”sparkrun updateThis updates core, reconciles compatible managed plugins, and refreshes recipe registries. Use sparkrun update --no-update-plugins for a local plugin compatibility audit without fetching plugin updates, or --no-update-registries to skip registry refresh. sparkrun setup update accepts the same options.
See Managed Plugins for separately installed integrations and What’s New in 0.4.0 for release highlights.
Requirements
Section titled “Requirements”- Python 3.12+
- SSH access to one or more NVIDIA DGX Spark systems
- Docker installed on target hosts (SSH user must be in the
dockergroup)