The retriever is often based on models like BERT (Bidirectional Encoder Representations from Transformers), which may effectively research and rank files primarily based on their relevance into the input https://www.juust.org/index.php/build-a-free-rag-system-with-n8n/2024/09/