Added postgresql sync and async checkpoint classes

[sharan9678]

I have added the postgresql sync and async checkpoints classes using psycopg3. To run the test we need to have a postgresql installed locally and then change the connection string in the both sync and async pregal test files.

[ajjjj8]

@sharanrajt do we have any estimate on when this can be merged?

[hinthornw]

Hi @sharan9678 , to avoid extra maintenance burdens in the langgraph package, we don't plan to add any additional integrations in the main langgraph package that add dependencies

[ootkin]

I was reviewing the Langgraph documentation and came across this note:

While this class does support asynchronous checkpointing, it is not recommended for production workloads due to limitations in SQLite's write performance. For production workloads, consider using a more robust database like PostgreSQL.

I agree with this recommendation, but the documentation does not provide guidance on using other production-ready databases (SQL, MongoDB, Redis, etc.).

I will check @sharan9678's code for further insights.

[sharan9678]

Hi @ootkin,

I had some similar requirements for PostgreSQL hence I wrote this code, But as of now langchain-ai provides support for PostgreSQL Checkpointer. Check the documentation at langchain-ai/langchain-postgres.

[flavioaltimari]

Could someone help on that? I need to build the checkpoint on postgresql, but lost here

[FilippoFlorindiKAI]

I have the same problem and I need to store checkpointers in a postgreSQL db. They also removed the checkpointer from the langchain-ai/langchain-postgres library.