Artifact Manifest
This section provides a comprehensive list of all components required for NVIDIA Cloud Functions (NVCF) Self-Hosted deployment for basic inference. Additional components are needed for Low Latency Streaming (Simulation).
Artifacts Overview
The following tables list all artifacts required for an inference-only self-hosted NVCF deployment, organized by category, with their container images, Helm charts, and other resources.
Early Access (EA) Version Policy
During Early Access, artifact versions are updated frequently. The versions shown for Infrastructure Components are stable references, but all other components should use the latest published version from NGC.
To find the latest versions:
First, ensure you have the latest version of the NGC CLI installed and configured.
Helm chart types
Rows marked Chart (OCI) are OCI-compliant charts stored in the NGC container registry. This means:
- Charts are pulled using
oci://URLs:helm pull oci://nvcr.io/0833294136851237/nvcf-ncp-staging/<chart-name> --version <version> - Charts are listed using the image registry command:
ngc registry image list - When mirroring to private registries (e.g., ECR), use container image tools like
skopeoorhelm push/pullwith OCI support
Rows marked Chart (HTTP) are traditional Helm repository charts, not OCI
URLs. In this manifest,
https://helm.ngc.nvidia.com/nvidia/omniverse/ddcs:5.0.0 means the chart
ddcs in the omniverse Helm repository
(https://helm.ngc.nvidia.com/nvidia/omniverse), at version 5.0.0. Add the
Helm repository and pull the chart by name and version, for example:
Some supporting components such as the GPU Operator, OpenBao, NATS, Cassandra, etc. can alternatively be pulled directly from public NGC Catalog or other public opensource repositories if desired.
Artifact Registry Paths
Infrastructure Components
Core infrastructure services including NATS for messaging, NATS auth callout support, Cassandra for data storage, and OpenBao for secret management.
Control Plane Components
Services that manage the NVCF platform including API gateway, deployment orchestration, invocation handling, LLM routing, and security services.
GPU Workload Components
Components that run on GPU nodes to manage function execution, including the NVCA operator and supporting containers.
Supporting Components
Additional utilities and helper services required for the platform, including the NVIDIA GPU Operator for GPU node management.
Reference Architecture Components
Optional components for the reference deployment architecture.
Observability Components
Optional example components for monitoring and observability. These are provided as reference implementations only and are not intended for production use. See self-hosted-example-dashboards for deployment instructions.
Container Caching Components
Optional components for accelerating container image pulls across all workload types.
Simulation Caching Components
Optional caching components for Low Latency Streaming (LLS) and simulation workloads, including shader caching, derived data caching, and USD content caching.
Storage API Components
Optional components for USD Storage API functionality used in simulation workloads.
Low Latency Streaming (LLS) Components
Components for Low Latency Streaming functionality.
Other Published Components
Additional components present in the current stack artifact manifest.
Deployment Resources
Helmfile and CLI resources for deployment.