Applies to SUSE AI Factory

4 What is the blueprint directory structure? #

While only the custom resource definition is mandatory, organizing files in a standardized directory structure ensures consistency across blueprints.

Partner blueprints are maintained in the SUSE partner-blueprints repository. Each blueprint should follow the directory structure described below.

blueprints/<ID>/
  <ID>-1-0-0.yaml        # REQUIRED — the Blueprint CR
  README.md              # Documentation and architecture details
  ATTRIBUTION.md         # Attribution details if derived from an upstream open source project
  clusterrepos/*.yaml    # Additional ClusterRepo definitions to apply before the CR

If the workload requires additional Kubernetes objects, store them in the same directory and document their deployment procedure in README.md.

Naming conventions:

Directory ID: Matches the value in metadata.labels[ai-factory.suse.com/blueprint-name].
CR file name: Uses the format <ID>-1-0-0.yaml. The metadata.name field uses hyphens (for example, my-demo-ollama-1-0-0), while spec.version and the version label use dots (such as 1.0.0).
Hardware variants: Maintain CPU and GPU variants in separate sibling directories (such as <NAME>-ollama and <NAME>-vllm), using displayName suffixes such as (Ollama, CPU) and (vLLM, GPU) and cross-linking them in their respective README.md files.