i would be curious if you are just using the vector database directly or using Pinecone Assistant, which does all the RAG pipeline pieces for you