I've been asked to ask the following during office hours: What is the granularity of the RAG in terms of how many document excerpts are provided? Is it variable or dependent on embedding match strength? How large are the excerpts? Are any of those parameters adjustable?