Skip to main content

vantage_sdk.workbench.nim_deployment

SDK helpers for cluster-scoped NVIDIA NIM deployment operations.

Classes​

NimDeploymentSDK()​

SDK for cluster-scoped NIM catalog + deployment operations.

Methods​

  • catalog(self, ctx, *, cluster_name: 'str', task: 'str | None' = None, publisher: 'str | None' = None, q: 'str | None' = None) -> 'ClusterServiceResponse': List NIM catalog entries, optionally filtered.
  • compatibility(self, ctx, *, cluster_name: 'str', catalog_id: 'str') -> 'ClusterServiceResponse': GPU compatibility targets for a catalog entry against cluster pools.
  • create(self, ctx, *, cluster_name: 'str', catalog_id: 'str', version: 'str', name: 'str', namespace: 'str | None' = None, profile: 'str | None' = None, platform: 'str' = 'standalone', precache: 'bool' = False, replicas_min: 'int' = 1, replicas_max: 'int' = 1, cache_size: 'str | None' = None, storage_class: 'str | None' = None, ingress_type: 'str' = 'public', ingress_auth: 'str' = 'user', env: 'dict[str, str] | None' = None, compute_pool: 'str | None' = None, pin_digest: 'bool' = False) -> 'ClusterServiceResponse': Deploy a NIM from the catalog (NIMService CR).
  • delete(self, ctx, *, cluster_name: 'str', name: 'str') -> 'ClusterServiceResponse': Delete a NIM deployment (NIMService + best-effort cache/PVCs).
  • get(self, ctx, *, cluster_name: 'str', name: 'str') -> 'ClusterServiceResponse': Get one NIM deployment by name.
  • list(self, ctx, *, cluster_name: 'str') -> 'ClusterServiceResponse': List the caller's NIM deployments.
  • versions(self, ctx, *, cluster_name: 'str', catalog_id: 'str') -> 'ClusterServiceResponse': List approved/blocked versions for a catalog entry (id = 'publisher/name').