Savvy's capabilities are essentially a typical application of RAG (Retrieval-Augmented Generation) architecture, retrieving document fragments first and then having a large language model generate responses.