Sample catalog

The repository ships several config/samples/ directories covering progressively more complex deployments. Apply them directly — each directory is self-contained and namespace-scoped.

Available samples

Directory	Contents	What it demonstrates
minimal/	PalenaGateway only	The smallest useful deployment — gateway with managed CNPG + Redis.
with-ui/	Gateway + Model + UI	Adds LibreChat on top of the minimal gateway.
with-observability/	Gateway + Model + Observability	Adds Langfuse tracing.
with-mcp/	Standalone PalenaMCPServer	Deploys the websearch MCP server with all sidecars. No gateway required.
full/	All five CRDs	Production-ish example: gateway + two models + UI with MCP refs + Langfuse + websearch.

Minimal

apiVersion: operator.palena.ai/v1alpha1
kind: PalenaGateway
metadata:
  name: production
  namespace: palena
spec:
  database:
    managed:
      instances: 1
      storageSize: 10Gi
  redis:
    managed:
      replicas: 1
      storageSize: 5Gi
  gateway:
    replicas: 1
    masterKey: { name: litellm-keys, key: master }
    saltKey:   { name: litellm-keys, key: salt }

With observability

apiVersion: operator.palena.ai/v1alpha1
kind: PalenaObservability
metadata:
  name: obs
  namespace: palena
spec:
  gatewayRef:
    name: production
  langfuse:
    web:
      replicas: 1
    worker:
      replicas: 1
    clickhouse:
      managed:
        replicas: 1
        storageSize: 50Gi
    blobStorage:
      provider: s3
      s3:
        region: us-east-1
        bucket: palena-langfuse-traces
        credentials:
          name: s3-credentials
          keys:
            accessKeyId: AWS_ACCESS_KEY_ID
            secretAccessKey: AWS_SECRET_ACCESS_KEY
    ingress:
      enabled: true
      host: langfuse.example.com
      className: nginx

Once applied, every request flowing through the gateway is automatically traced — Palena wires the Langfuse callback into the LiteLLMInstance and stores the seeded API keys in a generated Secret.

With MCP servers

A standalone websearch MCP server with all sidecars:

apiVersion: operator.palena.ai/v1alpha1
kind: PalenaMCPServer
metadata:
  name: websearch
  namespace: palena
spec:
  type: websearch
  replicas: 1
  websearch:
    searxng:
      replicas: 1
      engines: [duckduckgo, bing, wikipedia]
    scraper:
      chromiumEnabled: true
      chromiumReplicas: 1
      maxConcurrency: 10
    presidio:
      enabled: true
      mode: redact
    reranker:
      provider: flashrank

Register it from a PalenaUI:

apiVersion: operator.palena.ai/v1alpha1
kind: PalenaUI
metadata:
  name: chat
spec:
  gatewayRef: { name: production }
  mcpServerRefs:
    - name: websearch
  # ... rest of spec

LibreChat's librechat.yaml will be auto-regenerated with the MCP endpoint, and the Deployment rolls because the ConfigMap's SHA-256 is attached as a pod annotation.

Full stack

See config/samples/full/full.yaml in the repo — it wires all five CRDs together with realistic production settings (TLS via cert-manager, NetworkPolicies enabled, autoscaling, multiple models).