Could be a simpler setup than RAG for slow-changing documentation, especially for read-heavy cases.
No, that's one of the primary reasons for RAG.
Unless you have some evals showing that the previous results justifying RAG also apply to GPT-4o?
