LiteLLM

Description

Deploys a LiteLLM proxy that gives your services a single OpenAI-compatible API for calling multiple model providers. Other installations reference the endpoint output and authenticate with the master key to send requests.

Installs a single stateless Helm release in the litellm namespace -- no database needed. Comes preconfigured with two model routes: anthropic-default and openai-default. Bring your own API keys for each provider. You can swap the underlying model for either route (e.g. switch from claude-sonnet-4-6 to claude-opus-4-6) without changing any consuming service.

Enable "Expose Dashboard" to publish the proxy at <installation-name>.<environment-domain> with auto-provisioned TLS.

Configuration Inputs

This blueprint accepts 6 configurable inputs to customize the installation for your environment.

exposeDashboardboolean

Publish the LiteLLM proxy at a public HTTPS URL.

anthropicModelstring

Model string for the Anthropic route (format: anthropic/model-name). Your services call this as model name 'anthropic-default'.

openaiModelstring

Model string for the OpenAI route (format: openai/model-name). Your services call this as model name 'openai-default'.

anthropicApiKeystring

Anthropic API key for the anthropic-default model route.

openaiApiKeystring

OpenAI API key for the openai-default model route.

masterKeystring

Bearer token your services include when calling the proxy. Auto-generated if left blank.

Outputs

This blueprint exposes 2 outputs that other services can reference.

endpointstring

Cluster-internal URL for the LiteLLM proxy. Other installations reference this as their LLM API endpoint.

dashboardUrlstring

Public HTTPS URL for the LiteLLM proxy.