This is great! Reading the docs tickles my brain. Nice way to package up LLM functionality in a declarative way!