Installation
Prerequisites
Before choosing a deployment path, ensure these tools are installed:
| Tool | Version | Purpose |
|---|---|---|
| Helm | v3.19+ | Kubernetes package manager — Install |
| OpenShift CLI (oc) | Latest | Kubernetes CLI (includes kubectl) — Install |
| jq | Latest | JSON processor — Install |
| yq | Latest | YAML processor — Install |
| Python | 3.8+ | Deployment scripts — Install |
Additional prerequisites for specific paths
- Kind / NVKind: Docker installed and running
- NVKind: NVIDIA drivers + Docker GPU runtime (
nvidia-smimust detect your GPU) - Local (Lima/Minikube): Lima v1.2.1+ or Minikube, minimum 8GB RAM / 4 CPUs
- Cloud K8s / OpenShift: Existing cluster with kubectl/oc access configured
Not sure?
Start with Lima
Deployment Options
Local Deployment
For development and testing on your local machine:
| Deployment Type | Description | Guide |
|---|---|---|
| Local | Lima VM + Kubernetes on your machine | Deploy → |
Cluster Deployment
For production deployments on Kubernetes clusters:
| Deployment Type | Description | Guide |
|---|---|---|
| OpenShift Cluster | Red Hat OpenShift for enterprise | Deploy → |
| Kubernetes Cluster | Standard K8s cluster (GKE, EKS, AKS) | Deploy → |
| Kind | Kubernetes in Docker (local dev) | Deploy → |
| NVKind | Kind with NVIDIA GPU support | Deploy → |
After deployment
Once deployment is complete, return to First Steps to access your services, generate an API key, and run your first inference.