Hi, I'm trying to figure if there is an indexing/retrieval approach that lets me chunk documents, embed and store them into a vectordb while keeping track of the chunks that belongs to the same document. Then at retrieval time, return all chunks of the document if any of the chunks are within top_k in vector similarity.
I guess to some extent this sounds like the DocumentSummaryIndex. But I don't need the summary just the relations between chunks, and also DocumentSummaryIndex does not yet support embeddings?
