skills.rest
bcheng004
bcheng004COMMUNITY
Explore all skills by this author →
WebsiteGitHub

agent-memory

Persist agent conversation checkpoints and user preferences via Databricks Lakebase.

2|Updated Apr 6, 2026
Search Tags:#tool integration#checkpointing#langgraph#agent memory#user preferences#lakebase#long-term storage
One-click install
npx skills add https://github.com/bcheng004/agent-langgraph-advanced-workshop --skill agent-memory-bcheng004
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-memory
Source: https://github.com/bcheng004/agent-langgraph-advanced-workshop/tree/main/.claude/skills/agent-memory
Command: npx skills add https://github.com/bcheng004/agent-langgraph-advanced-workshop --skill agent-memory-bcheng004

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Agent responses lose useful context when conversation history and user-specific facts are not retained across requests, sessions, or users.

Core Features & Use Cases

  • Short-term memory (session context): Store conversation history within a session using AsyncCheckpointSaver, keyed by thread_id.
  • Long-term memory (user facts): Persist preferences and saved information across sessions using AsyncDatabricksStore, exposed via tools to get, save, and delete user memories.
  • Tool-first integration: Plug memory tools into your existing agent flow and pass the required store and user_id via RunnableConfig.

Quick Start

Add the memory dependency to your project, then integrate the memory tools and configure Lakebase so the agent can save and recall user memories using custom_inputs.user_id.

Frequently Asked Questions about agent-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I persist conversation history and user preferences across agent sessions?▼

Persistent agent memory is achieved by using AsyncCheckpointSaver for short-term conversation history keyed by thread_id and AsyncDatabricksStore for long-term user preferences keyed by user_id.

What's the best way to maintain user context in a long-running API-driven agent?▼

Maintaining user context in long-running API agents requires configuring memory tools that save and recall user facts, passing the store and user_id via RunnableConfig to preserve state across requests.

Does LangGraph support checkpointing for chat agents that need to remember prior interactions?▼

LangGraph supports checkpointing for chat agents by integrating AsyncCheckpointSaver, which stores session history by thread_id so agents can remember prior interactions within a conversation.

How do I configure Lakebase tables for storing long-term agent memory?▼

Configuring Lakebase tables for long-term agent memory requires executing store.setup() during initialization, enabling AsyncDatabricksStore to save, get, and delete user preferences across sessions.

Can I use custom user IDs to manage individual memory storage for different users?▼

Individual memory storage is managed by passing custom_inputs.user_id through RunnableConfig, allowing the agent tools to correctly associate saved preferences with specific users.

Do I need a separate database to store short-term conversation history and long-term user facts?▼

A single integrated agent flow manages both memory types: AsyncCheckpointSaver handles short-term conversation checkpointing and AsyncDatabricksStore manages long-term user facts.

Related Skills

Explore More Software Engineering Skills→
databricksdatabricks
183

agent-memory

Integrate long-term and short-term memory into LangGraph agents with Databricks Lakebase.

Official
Advanced
AshDaxAshDax

agent-memory

Persist user data across LangGraph agent sessions with Lakebase-backed memory tools.

Community
Advanced
caobingshengcaobingsheng
1

ai-agent-development

Automate end-to-end AI agent development and orchestration workflows.

Community
Advanced
victorlouvictorlou
4

agent-openai-memory

Store and retrieve multi-turn conversation history using OpenAI Agents SDK sessions.

Community
Intermediate