LiteLLM setup: providers, virtual keys and usage tracking

This tutorial introduces LiteLLM as an open source gateway with a unified API for multiple LLM providers. The presenter describes a free self-hosted option through the project's GitHub installation guide, then demonstrates deployment through Elestio, choosing a provider, region and service plan before signing in.

The walkthrough connects OpenAI with an API key and adds Google AI Studio with Gemini 3 Flash Preview. Credentials can be entered for a model or saved for reuse. Testing a connection does not save the configuration: the presenter explicitly reminds viewers to press Add Model.

Team settings control model access, spending budgets and token or request limits. The example uses a $100 budget with a selectable reset period; this is a demonstration setting, not a service price. Virtual keys let users make requests without receiving the underlying provider credentials, with separate permissions for LLM calls and management routes.

In the playground, the presenter uses a virtual key and switches between OpenAI and Gemini within one conversation. Usage analytics show requests and spending, with filters for teams, users, organizations and time ranges. This demonstration uses hosted provider APIs rather than showing how to run models locally. Agent and MCP server options receive only a brief mention, with viewers directed to the documentation.