langgraph-persistence

To save and load LangGraph state, you configure checkpointing to capture the graph state at every super-step, ensuring durable execution and enabling stateful AI applications. This allows the graph to resume from the last saved state.

LangGraph state persistence supports time travel by allowing you to browse checkpoint history. You can replay execution from a past state or fork to explore different outcomes without losing the original state history.

You manage isolated state using thread management features, which maintain separate state across different thread IDs. This ensures concurrent multi-threaded conversations remain independent and do not overwrite each other.

You can share user preferences and facts across conversations using the Long-Term Memory Store. This provides persistent storage that remains accessible across different sessions beyond standard checkpoint history.

Yes, production setups require database connectivity for LangGraph checkpointing. You need Python libraries such as psycopg2 or sqlite3 to ensure durable state persistence and reliable data storage.

Subgraph checkpointer scoping controls persistence behavior for subgraphs using different checkpointer modes. This allows you to isolate or share state within nested graph structures for complex application workflows.