But I think the eventual goal is that documentations won't even be needed. LLM should just itself understand the nuances of frameworks by analyzing their codebase.