What problem does it solve?
LangGraph runs often lose conversational context and intermediate progress between invocations, making multi-turn experiences unreliable and forcing developers to rebuild state management manually.
Core Features & Use Cases
- Checkpointing durable graph state: Save and restore graph execution state at each super-step using a configurable checkpointer.
- Thread-scoped conversation continuity: Use a
thread_id to separate and persist distinct dialog histories without mixing users/sessions.
- Cross-thread long-term memory with Store: Persist user-level preferences and facts beyond a single thread so different conversations can share context.
- Subgraph checkpointer scoping: Control persistence behavior when compiling subgraphs to enable interrupts, multi-turn memory, and safe parallelization patterns.
Example use cases:
- A customer support agent that remembers a user’s ongoing conversation even after restarts.
- A travel assistant that retains long-term preferences (e.g., budget, dietary needs) across multiple sessions.
- A system that resumes or forks execution from checkpoint history to implement “time travel” debugging and recovery.
Quick Start
Use the langgraph-persistence skill to add checkpoint persistence to your graph by compiling with an appropriate checkpointer and always passing a configurable.thread_id when invoking the graph.
