Pro tip: you can just feed the LLM an OpenAPI spec and it will work just as well.