Palena OperatorEnterprise LLM platform. Declarative Kubernetes.

Why Palena? ​

Running an enterprise LLM platform means stitching together a gateway, a database, a cache, a chat UI, a trace collector, and a pile of MCP servers. Palena does the stitching.

Write a PalenaGateway CR, and the operator creates:

• A CloudNativePG cluster for LiteLLM's backend state
• A Redis instance for rate limiting and routing
• A LiteLLMInstance CR with database/Redis references pre-wired
• NetworkPolicies that only allow the communication paths you need

Layer on the other four CRDs and you get models, a chat UI, Langfuse tracing, and a websearch MCP server — every piece owned, every connection pre-wired.

Architecture at a glance ​

Quick install ​

operator-sdk run bundle ghcr.io/palenaai/palena-operator-bundle:v0.1.0

helm install palena-operator deploy/charts/palena-operator \
  --namespace palena-system \
  --create-namespace

make install
make deploy IMG=ghcr.io/palenaai/palena-operator:v0.1.0

See the installation guide for the full walkthrough.

What does "Palena" mean? ​

"Palena" is Hawaiian for boundary, limit, border — the operator defines the perimeter inside which your AI platform runs. Every resource is owned, every credential is scoped, every network path is explicit.