langgraph-persistence

Compile the graph with a checkpointer and always pass a thread_id in the config. Use InMemorySaver for development or PostgresSaver for production, then invoke with {"configurable": {"thread_id": "session-1"}} so state persists across calls.

PostgresSaver is the production-ready checkpointer, created via PostgresSaver.from_conn_string with a one-time setup() call to create tables. InMemorySaver loses data on restart and SqliteSaver is only suited for local development.

Call get_state_history (Python) or getStateHistory (TypeScript) with the thread config to list past checkpoints. Resume from one by invoking with None/null and the past state's config, or fork by calling update_state first.

State is not persisted when thread_id is missing from the invocation config. Always pass {"configurable": {"thread_id": "..."}} on every invoke call so the checkpointer can save and reload the thread's state.

No. A subgraph compiled with checkpointer=True cannot be called multiple times within a single node because the calls write to the same checkpoint namespace and conflict. Use checkpointer=None for parallel invocations or wrap different subgraphs with unique node names.

Use a Store such as InMemoryStore compiled alongside the checkpointer. Access it inside nodes via the Runtime parameter (runtime.store) and key entries by user namespace so preferences persist across all threads.