Agent state and memory
Lakebase Autoscaling is the latest version of Lakebase, with autoscaling compute, scale-to-zero, branching, and instant restore. For supported regions, see Region availability. If you are a Lakebase Provisioned user, see Lakebase Provisioned.
AI agents need persistent storage to maintain context across turns and sessions. Lakebase Autoscaling provides a fully managed Postgres backend for storing agent state and memory, integrating natively with Databricks authentication and scaling automatically with your workload.
Short-term vs. long-term memory
Short-term memory | Long-term memory |
|---|---|
Captures context within a single conversation session using thread IDs and checkpointing. Lets agents answer follow-up questions with awareness of earlier turns. | Extracts and stores key insights across multiple conversations. Enables personalized responses based on past interactions. Builds a user knowledge base that improves over time. |
You can implement either or both memory types in the same agent.
Deployment options
Lakebase-backed agent memory is supported on two Databricks deployment targets:
Databricks Apps: Deploy agents as interactive applications with short-term or long-term memory, using LangGraph checkpointers or the OpenAI Agents SDK. Databricks handles authentication between the app and Lakebase automatically. See AI agent memory.
Mosaic AI Model Serving: Deploy agents to Model Serving endpoints with Lakebase-backed checkpoints. Supports LangGraph time travel to resume or fork conversations from any checkpoint. See AI agent memory (Model Serving).
Implementation
For full setup instructions, app templates, and notebook examples, see: