agent-memory

To add long-term memory to LangGraph agents, use the memory_tools() factory to integrate an AsyncDatabricksStore. This persists user data and conversation history across sessions by scoping memories with a provided user_id.

Yes, you can use Lakebase to store user preferences during streaming workflows by wiring memory_tools() with an AsyncDatabricksStore. This requires passing a user_id in requests to properly scope and save the memories.

The available memory functions for managing persistent agent context are get_user_memory, save_user_memory, and delete_user_memory. These are provided by the memory_tools() factory to retrieve, store, and remove user data.

Yes, you need a user_id to scope memories in a Lakebase-backed store. Passing a user_id in requests ensures that the AsyncDatabricksStore correctly partitions and retrieves long-term memory for individual users.

Your LangGraph agent fails to recall context across sessions if long-term memory is not configured. You must wire memory_tools() with a compatible store like AsyncDatabricksStore and pass a user_id to persist data.

The best way to integrate conversation history checkpoints into existing LangGraph agents is by applying memory_tools() with a Lakebase-backed store. This enables seamless recall of user preferences and history without disrupting streaming workflows.