Skip to main content

Kubernetes

Libre WebUI ships a Helm chart under helm/libre-webui.

Install​

helm install libre-webui oci://ghcr.io/libre-webui/charts/libre-webui

The default chart deploys Libre WebUI with persistent storage and a bundled Ollama service.

Access Locally​

kubectl port-forward svc/libre-webui 8080:8080

Open http://localhost:8080.

External Ollama​

Use an existing Ollama endpoint:

helm install libre-webui oci://ghcr.io/libre-webui/charts/libre-webui \
  --set ollama.bundled.enabled=false \
  --set ollama.external.enabled=true \
  --set ollama.external.url=http://my-ollama:11434

Secrets​

Set a stable JWT secret and encryption key for production. Use Kubernetes Secrets rather than plaintext values files.

kubectl create secret generic libre-webui-secrets \
  --from-literal=JWT_SECRET="$(openssl rand -hex 64)" \
  --from-literal=ENCRYPTION_KEY="$(openssl rand -hex 32)"

Add provider API keys the same way when you want deployment-wide plugin credentials.

Persistence​

Keep the Libre WebUI data PVC and Ollama model PVC on persistent storage. Back up the Libre WebUI data volume and the encryption key together.

Ingress​

For public access, configure ingress with HTTPS and set backend origins:

BASE_URL=https://your-domain.example
CORS_ORIGIN=https://your-domain.example

OAuth callback URLs must match the public domain.

Resource Planning​

For local Ollama inside the cluster, schedule the Ollama pod on nodes with enough memory and GPU capacity for the models you plan to run. If your cluster already has a dedicated Ollama or inference service, external Ollama is usually simpler.