I have been experimenting with the langgraph-checkpoint-postgres library; I’m pretty new to agent development so I haven’t tried anything too exciting yet but so far it works great! My only gripe as a Django developer is that it adds tables to the database that aren’t managed by Django.
I can (and will, for now) work around this by creating a migration that performs a RunPython operation instantiating a PostgresSaverand running the .setup() method on it, but I’m considering writing a library similar to langgraph-checkpoint-postgres that relies on Django and provides models, uses ORM for working with the database, etc. The concept I have in mind is to package the checkpointer as a Django app with models, migrations, and versions of PostgresSaver, AsyncPostgresSaver, etc. that use those models. Could probably do something for other databases too.
I’m putting this message out to see if there’s anyone else out there who’d be interested in something like this. If I’m alone I’ll probably just go with the workaround above but if others would find it useful, I’d love to contribute to LangGraph.