Context: I’m trying to implement an advanced RAG pipeline that uses Auto-merging Retriever (aka Parent Document Retriever) against specific VectorDB (for example, Pinecone).