Token Max parameter This impacts the ReduceDocumentChain, which can often be used to power the ConversationalRetrievalChain This parameter controls how large of chunks (in token length) to combine documents into before reducing Docs:

→ View original post on X — @hwchase17