FreeLLMAPI setup with Kilo Code and provider API keys

The tutorial sets up FreeLLMAPI as a self-hosted router for external model providers. It combines connected provider accounts behind an OpenAI-compatible endpoint and API key. The speaker describes routing around provider limits and switching providers when an endpoint is slow or unavailable. Running the router on your computer still sends model requests to those providers.

Setup starts with cloning the repository and installing dependencies with npm. The application requires a random 32-byte encryption key in an environment variable. The presenter generates it with Node.js in PowerShell, noting that commands can differ between terminals, then starts the development server and creates a dashboard account.

The walkthrough connects Navy AI and Mistral using their API keys, checks provider health, and tests responses in the playground. The catalog shows remaining quotas and filters for speed, context length, and model capabilities. The speaker's Mistral response feels faster in this test; that is not a general performance benchmark.

For the coding assistant demonstration, the presenter configures Kilo Code in VS Code with a custom provider, the router's API key, and a local base URL using port 5000 and V1. Other setups may use a different port. Mistral Large answers a message and explains the project. The closing caveat is that providers can restrict use of their free tiers.