NVIDIA Network Operator v26.4.0

Automation and CI/CD

Note

Use this when: you are wiring Launch Kit into a CI/CD pipeline, an automation framework, or an AI agent that needs structured output and exit-code-driven retry logic.

Launch Kit is designed for both interactive use and automated pipelines. Every subcommand supports a JSON output mode, well-defined exit codes, and a programmatic capability discovery endpoint.

Use --output json for machine-readable output. In JSON mode, structured data goes to stdout and human-readable logs go to stderr:

Copy
Copied!
            

l8k generate --user-config ./cluster-config.yaml \ --fabric ethernet --deployment-type sriov --multirail \ --save-deployment-files ./deployments \ --output json --yes 2>/dev/null | jq .

  • --yes / -y — auto-confirm all prompts (implied by --output json)

  • --quiet / -q — suppress informational output

Code

Meaning

0 Success.
1 General error.
2 Validation error (invalid flags or configuration).
3 Cluster error (API unreachable, discovery failed).
4 Deployment error (apply failed).
5 Partial success (discovery completed but deployment failed).

In JSON mode, errors include structured fields (code, category, transient, suggestion) to help automation decide whether to retry or fix input. transient: true indicates a temporary condition (e.g., API unreachable) where retry is appropriate; transient: false indicates a permanent failure (e.g., invalid flag) requiring input changes.

AI agents and automation tools can programmatically discover Launch Kit’s capabilities:

Copy
Copied!
            

l8k schema

Outputs a JSON description of available commands, phases, fabrics, deployment types, flags, exit codes, and output formats. Tools that need to know what l8k can do (without parsing --help text) use this endpoint.

Previous Cluster Topology Presets
Next Troubleshooting
© Copyright 2025-2026, NVIDIA. Last updated on Jun 14, 2026