Last night I finally realized the embeddings model you choose has a great impact on your RAG workflow and the quality of the answers returned from the LLM. A lot of testing today and tomorrow.