Whereas the way RAG inside of Typesense works is we first take the query and do a semantic search within your dataset, take the top N items and send it to the LLM to formulate a response