RAG is designed more for text-based semantic matching. For structured querying like this, I would recommend building your own LLM integration outside of Typesense where, you take the question and ask the LLM to formulate a query to send to Typesense.