Skip to main content
Version: Next

KServe Installation

This guide covers installation of the KServe controller for predictive AI model serving.

Prerequisites​

Before installing KServe, ensure dependencies are installed based on your deployment mode:

Deployment Modes​

KServe supports two deployment modes:

ModeDescription
KnativeServerless deployment with Knative
StandardRaw Kubernetes deployments using base Kubernetes features

Installation Methods​

Method 1: Kustomize​

Knative Mode​

# Clone KServe repository
git clone https://github.com/kserve/kserve.git
cd kserve

# Install KServe with all components
kubectl apply -k config/overlays/standalone/kserve

For addon installation (when LLMIsvc is already installed):

# Install only KServe component (no base resources, reuses existing namespace)
kubectl apply -k config/overlays/addons/kserve

Standard Mode​

# Set deployment mode to Standard
# Edit the inferenceservice ConfigMap
vi config/default/configmap/inferenceservice.yaml

# Change the deploy section to:
# deploy: |
#   {
#     "defaultDeploymentMode": "Standard"
#   }

# Apply
kubectl apply -k config/overlays/standalone/kserve

Install ClusterServingRuntimes​

# Install all runtimes
kubectl apply -k config/runtimes

Method 2: Helm​

Install CRDs​

First, install the KServe CRDs:

# Using OCI registry (recommended)
helm install kserve-crd oci://ghcr.io/kserve/charts/kserve-crd \
  --version v0.17.0-rc1 \
  --namespace kserve \
  --create-namespace

# Or using local charts
helm install kserve-crd ./charts/kserve-crd \
  --namespace kserve \
  --create-namespace

Install KServe Resources​

Knative Mode:

helm install kserve-resources oci://ghcr.io/kserve/charts/kserve-resources \
  --version v0.17.0-rc1 \
  --namespace kserve \
  --set kserve.controller.deploymentMode=Knative

Standard Mode:

helm install kserve-resources oci://ghcr.io/kserve/charts/kserve-resources \
  --version v0.17.0-rc1 \
  --namespace kserve \
  --set kserve.controller.deploymentMode=Standard

Addon Installation (when LLMIsvc is already installed):

helm install kserve-resources oci://ghcr.io/kserve/charts/kserve-resources \
  --version v0.17.0-rc1 \
  --namespace kserve \
  --set kserve.createSharedResources=false

Method 3: Installation Scripts​

Quick Install (All-in-One)​

Knative Mode:

cd kserve

# Install dependencies + KServe
./hack/setup/quick-install/kserve-knative-mode-full-install-helm.sh

# Or use with-manifest version (no clone needed, includes embedded manifests)
./hack/setup/quick-install/kserve-knative-mode-full-install-helm-with-manifest.sh

Standard Mode:

cd kserve

# Install dependencies + KServe
./hack/setup/quick-install/kserve-standard-mode-full-install-helm.sh

# Or use with-manifest version (no clone needed, includes embedded manifests)
./hack/setup/quick-install/kserve-standard-mode-full-install-helm-with-manifest.sh

Configuration Helm Options​

For detailed configuration options including deployment mode, gateway settings, resource limits, and runtime configurations, see the kserve-resources Helm Chart README.

Uninstallation​

Helm​

# Remove resources
helm uninstall kserve-runtime-configs -n kserve
helm uninstall kserve-resources -n kserve
helm uninstall kserve-crd -n kserve

# Remove namespace
kubectl delete namespace kserve

Kustomize​

# Remove KServe
kubectl delete -k config/overlays/standalone/kserve

Scripts​

# Uninstall using script
./hack/setup/quick-install/kserve-knative-mode-full-install-helm.sh --uninstall

# Or individual script
UNINSTALL=true ./hack/setup/infra/manage.kserve-helm.sh

Next Steps​