Usage guidelines with samples especially for ai_context field

[shubhambatlan]

for ai_context context field, it’s not clear whether all the available sub-fields are available at each level. e.g. if I add instructions for a given field or derived metric, how does that get used by the LLM?
perhaps a sample usage guideline on how to build agents with the OSI would be a good addition to the repo