langgraph-persistence

To maintain LangGraph state across invocations, initialize your graph with a persistent store and configure both the checkpointer and store settings. This ensures durable execution and conversation history tracking for production-level deployments.

Checkpointing in LangGraph is the mechanism for durable execution of the graph state through different checkpoints. You need it when building applications that require conversation state persistence across user sessions or during production deployments.

Yes, you can isolate conversation history using thread management features. This ensures isolated state handling for each thread, maintaining conversation privacy and preventing cross-contamination between different user interactions.

You manage subgraph scope settings by configuring different persistence behaviors for subgraphs based on various use cases. This allows you to handle parallelism in stateful subgraphs while maintaining appropriate state isolation and retention.

Yes, LangGraph supports time travel by allowing you to browse and interact with the graph's state history. This feature lets you review previous checkpoints and resume execution from specific historical states.

You implement long-term memory using the cross-thread memory storage feature. This provides persistent storage for user preferences and facts across different conversations, requiring specific store configurations during graph initialization.