We introduced the ParentDocumentRetriever last week to strike a balance between: – using small chunks during indexing
– passing larger chunks to the LLM Great overview and diagram by @clusteredbytes

→ View original post on X — @hwchase17,