Choose a blueprint

For developers and maintainers selecting a workflow from the configured catalog or a local package. This how-to covers discovery, review, preflight, and launch; it does not certify catalog contents or output quality. Sources are SDK blueprint_source.py and CLI blueprint resolution/launch tests. Catalog availability is version-dependent.

Discover the active catalog

From any directory with the CLI installed:

mn blueprint list
mn blueprint show <blueprint-id>

Replace <blueprint-id> with an ID returned by the list command. Review the package README, inputs, execution runner, model/service requirements, dependency versions, network access, and output actions. Use provided sample inputs first.

The catalog is configured by MN_BLUEPRINT_SOURCE, MN_BLUEPRINT_REPO, and MN_BLUEPRINT_LOCAL; see Environment Variables. A separately checked-out catalog is not assumed to exist beneath this runtime workspace.

Preflight a local package

From the directory containing your reviewed blueprint folder:

mn blueprint validate ./my-blueprint
mn blueprint doctor ./my-blueprint

Replace ./my-blueprint with the actual path. Both commands must complete without blocking errors before normal launch. Validate checks the package and requirements; doctor diagnoses runtime prerequisites. Neither proves output correctness or authorization for external actions.

Launch and inspect

Launching executes the package's code and can prepare dependencies and resources. Review files, credentials, and external actions first.

mn blueprint run ./my-blueprint --detached

For a catalog-managed package, use its ID instead of a path:

mn blueprint run <blueprint-id> --detached

Record the returned public run ID, then follow Quickstart to inspect results and cancel unfinished work. Preserve evidence before deleting run data.