Hi,
I’d like to contribute a new library libs/checkpoint-qdrant/ that adds
Qdrant as a backend for both the LangGraph checkpoint saver and the
long-term memory store.
Motivation: Qdrant is widely used in LangGraph-based apps for vector
search. Users who already run Qdrant in their stack currently have to
introduce a second database (Postgres or SQLite) just for checkpointing.
Having a native Qdrant implementation removes that friction.
Proposed API:
QdrantSaver/AsyncQdrantSaver— checkpoint saver
(langgraph.checkpoint.qdrant)QdrantStore/AsyncQdrantStore— memory store with optional vector
similarity search, namespace filtering, and TTL
(langgraph.store.qdrant)
The implementation follows the same structure and semantics as
checkpoint-postgres and checkpoint-sqlite. I have a working
implementation ready with 40 tests passing (ruff + mypy clean, tested
against both in-memory and real Docker Qdrant).
Happy to open a PR as soon as this is approved. Let me know if you’d
like me to adjust anything in the API design.
Current implementation here: GitHub - gianlucaperna/langgraph at feat/checkpoint-qdrant · GitHub
Thanks!