LiteLLM Operator

LiteLLM Operator provides declarative, reconciliation-based management for LiteLLM instances within Kubernetes environments, eliminating the need for manual Helm installations. Key features include: * Declarative custom resources for LiteLLM components * Bidirectional synchronization between Kubernetes CRDs and the Admin UI * Comprehensive management for models, teams, users, and API keys * Production-ready deployment features like HPA, PDB, and NetworkPolicy * Multiple installation methods, including OperatorHub/OpenShift support This operator manages LiteLLM instances, models, teams, users, and API keys as native Kubernetes custom resources, ensuring full GitOps compatibility. It features bidirectional configuration synchronization, allowing changes via CRDs or the Admin UI to be kept in sync with flexible policies for unmanaged resources (preserve, prune, or adopt). The system also includes SSO-aware team management with various modes to prevent accidental destruction of SSO-provisioned memberships. For enhanced security and ease of use, virtual keys are automatically generated via the LiteLLM API and stored in Kubernetes Secrets with owner references for proper garbage collection. The operator is designed for production readiness, offering out-of-the-box support for Horizontal Pod Autoscalers (HPA), Pod Disruption Budgets (PDB), NetworkPolicies, health probes, security contexts, resource limits, and database migrations. LiteLLM Operator is ideal for engineering teams and platform administrators who deploy and manage web gateway instances on Kubernetes, providing a robust and scalable solution for orchestrating their deployments. It supports various installation methods, including OLM bundles for OperatorHub and OpenShift, a Helm chart for vanilla Kubernetes, and direct Makefile-based installation.