Hi LangChain community,
LangChain Support suggested I share this here to route it to the right place.
I’ve built a standalone Time Checkpoint / Memory Folding Engine demo that behaves like an agent-memory backend with:
-
Deterministic time checkpoints (folded state snapshots over time)
-
Restart continuity (stop the process, restart, continue from the same truth state)
-
SHA-256 receipts + audit trail (verifiable artifacts written to disk)
-
Two integration shapes: library mode (direct calls) or local HTTP service (simple endpoints like state/checkpoint/restore)
This is packaged as a turnkey Windows/Python demo intended for private technical evaluation (not a public code drop). I’m happy to share an evaluator checklist and provide the demo pack privately to the appropriate LangGraph/LangChain contact.
Question: Who is the best person/team to evaluate this as a LangGraph memory provider / checkpoint backend (checkpoint/restore/audit) for a confidential review?
Preferred contact: DM me here, or email ristuben.tech@gmail.com.
Thanks,
John