Chain of Thought Prompts discussion here mentions interesting techniques and work arounds since you’re not allowed to do/advised against doing CoT on a few services
Coding
Still having issues with the model to model validation the documentation recommends
then when you use it you can write in it’s integrated AI chat “@” and it gives you a pop up menu. And in it you can choose “Docs” and “Add doc”. Then you fill in an url for Docs you would like your AI to use when fixing your code, like "https://docs.pydantic.dev/latest/"
that gets your custom docs indexed. After that you keep on chatting “could you fix my <code> pydantic error <error> according to @pydantic docs”?
when you tag the specific docs you just added cursor goes in and chooses which subpages to include. But in case it does poor job you can always surf your favorite doc page and say “fix my code according @https://docs.pydantic.dev/latest/concepts/validators/” and after ‘@’ and pasting the link you get a pop-up where you press enter for “Add link”