AsyncShallowRedisSaver Not Overwriting Checkpoint Blobs and Writes

[FedericoDellOrso]

Hello,
I’m using AsyncShallowRedisSaver in my LangGraph project, and I’ve noticed that checkpoint_blob and checkpoint_writes are not being overwritten in the database. Instead, they keep stacking up during graph execution. The checkpoint itself appears to be fine, but these leftover blobs and writes are not referenced in the final checkpoint.

I don’t think this is the intended behavior, as the database keeps filling up with unnecessary items. I also noticed that while there is a _make_shallow_redis_checkpoint_key method to generate the key for the checkpoint, there doesn’t seem to be a shallow implementation for blobs and writes.

Am I missing something? Will this be addressed in a future update, or should I open a PR to fix it?
Thanks!

[bsbodden]

@FedericoDellOrso confirmed! Thanks. We're working on a PR now.

[bsbodden]

@FedericoDellOrso fixed, can you test it?

[FedericoDellOrso]

Hi, i've tested it but sadly i can still see all the previous blobs in redis, i don't see any difference from the previous version.