ReMe: An Advanced Memory Management Kit for AI Agents

Introduction

ReMe, short for "Remember Me, Refine Me," is a powerful Memory Management Kit for AI Agents. This Python-based, local-first memory layer is designed to transform conversations and resources into file-based long-term memory. With over 3,200 stars and 280 forks on GitHub, ReMe provides a robust solution for agents to continuously index, link, and consolidate information for future recall. Its core philosophy revolves around treating memory as editable Markdown files, fostering a self-evolving knowledge base for intelligent agents.

Installation

ReMe requires Python 3.11 or newer. You can easily install it using pip:

pip install "reme-ai[core]"

For more advanced setups or to install from source, please refer to the official documentation.

Examples

Getting started with ReMe is straightforward. After installing and starting the service, you can quickly write, search, and read memory nodes.

First, start the ReMe service:

reme start

Then, write a memory node:

reme write \
  path=digest/wiki/quick-start-demo \
  name="Quick Start Demo" \
  description="A first ReMe memory node" \
  content="# Quick Start Demo\n\nReMe stores agent memory as readable Markdown.\n\nRelated: [[digest/wiki/memory-as-file.md]]"

Search for relevant memory:

reme search query="agent memory markdown" limit=5

And read a specific memory node:

reme read path=digest/wiki/quick-start-demo start_line=1 end_line=20

This demonstrates how ReMe stores agent memory as readable Markdown files, complete with frontmatter and wikilinks.

Why Use ReMe?

ReMe offers several compelling reasons for integration into your AI agent projects:

• Memory as File: It stores memory as Markdown files with frontmatter and wikilinks, making it directly readable and writable by both users and agents.
• Self-evolving Knowledge Base: Features like Auto Memory, Auto Resource, and Auto Dream progressively transform conversations and resources into long-term memories, automatically building wikilink relationships.
• Progressive Hybrid Search: ReMe combines wikilinks, BM25, and embeddings for hybrid retrieval, ensuring comprehensive search capabilities across keyword matching, semantic recall, and relationship expansion.
• Agent-friendly Integration: With SKILL.md and CLI integration, ReMe simplifies how different agents can read, write, maintain, and reuse memory.
• Diverse Use Cases: From powering personal assistants and coding agents to building LLM wikis and supporting self-evolving agents, ReMe provides a versatile memory layer. Its design philosophy emphasizes capturing raw dialogs and resources, then refining them into long-term preferences, reusable experience, and valuable knowledge, all while keeping the result editable by humans and agents.

Links

Explore ReMe further through these official resources:

• GitHub Repository: agentscope-ai/ReMe
• Official Documentation: ReMe Documentation
• PyPI Project: reme-ai on PyPI