LiteLLM

LiteLLM is a proxy that exposes one OpenAI-compatible API and routes each request to a configured model, so callers don't need to know which provider is behind a given model name.

Runs in the open-webui namespace, routing named models to the Anthropic API with the key read from a Kubernetes secret, and storing its model configuration and spend logs in its own PostgreSQL database.

A single routing layer means swapping or adding a model is a config change in one place instead of touching every client that calls it.