langgraph-persistence

To persist LangGraph state across conversations, you use a checkpointer with a unique thread_id to isolate and save state per session. This enables durable execution and ensures the conversation history remains accessible for later resumption.

Yes, you can share cross-thread memory in LangGraph by integrating a Store. The Store allows runtime access to shared user preferences or facts, enabling multiple isolated threads to access and update the same persistent data.

You resume or fork a LangGraph workflow by initializing the graph with a specific thread_id and referencing a previous checkpoint. This allows you to navigate state history and branch execution from prior conversation states.

You need a checkpointer for thread-scoped state history and execution durability, while a Store is needed for cross-thread memory sharing. Use a checkpointer to isolate conversations and a Store to access shared facts across multiple threads.

Yes, LangGraph persistence supports isolating state for subgraphs using checkpointers and thread_id. This allows you to maintain independent execution states and histories for nested workflows within the overall graph structure.

Using thread_id for state management isolates history to a single conversation thread, preventing direct memory sharing between separate threads. To overcome this limitation and share user facts globally, you must integrate a Store for cross-thread memory access.