How do I ensure long-term persistence doesn't reset on restart?

For production environments, use the PostgresStore instead of InMemoryStore. InMemoryStore is lost when the application process restarts, while PostgresStore keeps data in a permanent database.

What is the difference between StateBackend and StoreBackend?

StateBackend provides ephemeral storage that exists only within a single thread and is lost once the thread ends. StoreBackend provides long-term persistence that allows data to be shared and accessed across different threads and sessions.

Can I route different file paths to different storage engines?

Yes, using the CompositeBackend you can define routing rules. For example, you can route '/tmp/' to an ephemeral StateBackend while routing '/memories/' to a persistent StoreBackend.

Is FilesystemBackend safe for production web servers?

No, FilesystemBackend is designed primarily for local development and CLI tools. For web servers, you should use StateBackend or a sandboxed environment to prevent unauthorized access to the host's filesystem.

Deep Agents Memory

Name: Deep Agents Memory
Author: langchain-ai

bylangchain-ai

•

534

•

データサイエンスとML

Manages short-term and long-term memory for AI agents through pluggable storage backends and secure filesystem access.

Deep Agents Memory provides a comprehensive framework for managing state and persistence within AI agent workflows. It offers several specialized backends—including StateBackend for ephemeral thread-level memory, StoreBackend for cross-session persistence, and FilesystemBackend for direct disk access. By utilizing a CompositeBackend, developers can implement sophisticated routing logic to ensure sensitive data remains ephemeral while core memories persist across multiple threads. This skill is essential for building complex agents that need to retain context, follow long-running instructions, or interact with local codebases securely.

主な機能

01Integrated FilesystemMiddleware providing ls, read, write, and grep tools

02Cross-thread memory persistence using StoreBackend and PostgresStore

03Human-in-the-loop safety interrupts for filesystem modifications

04Pluggable backend architecture for ephemeral or persistent storage

05Flexible path-based routing via CompositeBackend for hybrid storage

06534 GitHub stars

ユースケース

01Developing local CLI agents that require restricted access to specific project directories

02Building AI assistants that retain user preferences and history across different chat sessions

03Creating multi-step agentic workflows that use temporary scratchpads for intermediate data

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add langchain-ai/langchain-skills deep-agents-memory

For use in Claude.ai and ChatGPT

主な機能

01Integrated FilesystemMiddleware providing ls, read, write, and grep tools

02Cross-thread memory persistence using StoreBackend and PostgresStore

03Human-in-the-loop safety interrupts for filesystem modifications

04Pluggable backend architecture for ephemeral or persistent storage

05Flexible path-based routing via CompositeBackend for hybrid storage

06534 GitHub stars

ユースケース

01Developing local CLI agents that require restricted access to specific project directories

02Building AI assistants that retain user preferences and history across different chat sessions

03Creating multi-step agentic workflows that use temporary scratchpads for intermediate data