Comment on LLM function calls don't scale; code orchestration is simpler, more effectiveparentComments−never_inline1y1. This is using a lot of input context in formating as JSON and escaping a Json inside already a JSON.Isn't it a model problem that they don't respect complex json schemas?
Comments
Isn't it a model problem that they don't respect complex json schemas?