That use case is very LLM unfriendly, because you want very specific and precise structure of output, but the LLM knows lots of stuff that is absolutely not wha

That use case is very LLM unfriendly, because you want very specific and precise structure of output, but the LLM knows lots of stuff that is absolutely not what you want at all.