After a few iterations, we found the best way to do that was to build a document store into LangGraph This comes with:
– Flexible namespacing (to support multiple types of memory)
– JSON Document Storage
– Content-Based Filtering

→ View original post on X — @hwchase17,