agentscope-ai/ReMe
ReMe: Memory Management Kit for Agents - Remember Me, Refine Me.
Python · ★ 2,049 · Fork 155 · 在 GitHub 查看
在手机上浏览源码、订阅每日 AI 趋势日报与项目架构深度解读:获取 CopoHub 客户端(iOS & HarmonyOS)→
GitHub Trending 上榜记录
agentscope-ai/ReMe 已在 GitHub Trending 每日榜单上榜 4 天,最高排名第 6,首次上榜 2026-03-03,最近一次上榜 2026-03-06。
| 日期 | 排名 | Star | 当日增长 |
|---|---|---|---|
| 2026-03-06 | #10 | 1,918 | +194 |
| 2026-03-05 | #10 | 1,781 | +195 |
| 2026-03-04 | #6 | 1,593 | +346 |
| 2026-03-03 | #6 | 1,243 | +27 |
同期上榜的项目
- moeru-ai/airi – 💖🧸 Self hosted, you-owned Grok Companion, a container of souls of waifu, cyber livings to bring them into our worlds, wishing to achieve Neuro-sama's altitude. Capable of realtime voice chat, Minecraft, Factorio playing. Web / macOS / Windows supported.
- QwenLM/Qwen-Agent – Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc.
- KeygraphHQ/shannon – Shannon Lite is an autonomous, white-box AI pentester for web applications and APIs. It analyzes your source code, identifies attack vectors, and executes real exploits to prove vulnerabilities before they reach production.
- microsoft/hve-core – A refined collection of Hypervelocity Engineering components (instructions, prompts, agents, and skills) to start your project off right, or upgrade your existing projects to get the most out of all Copilots
- Ed1s0nZ/CyberStrikeAI – CyberStrikeAI is an AI-native security testing platform built in Go. It integrates 100+ security tools, an intelligent orchestration engine, role-based testing with predefined security roles, a skills system with specialized testing skills, and comprehensive lifecycle management capabilities.
- aquasecurity/trivy – Find vulnerabilities, misconfigurations, secrets, SBOM in containers, Kubernetes, code repositories, clouds and more
ReMe 项目深度剖析与商业评估报告
ReMe 项目深度剖析与商业评估报告
1. 核心定位与商业落地价值
ReMe 解决的是 AI Agent 生态中一个被严重低估的根问题:长期记忆的可审计性与持续演化能力。当前主流 Agent 框架(LangChain、AutoGen、Claude Code)普遍采用向量数据库作为记忆载体,但向量检索本质上是"黑盒相似度匹配",无法回答"这个记忆从哪来、如何被修正、为什么被召回"三个关键问题。ReMe 通过"Memory as File"的设计哲学,将记忆降维为可读、可编辑、可审计的 Markdown 文件,从根本上切入了企业级 AI 应用对合规性、可解释性、人工干预能力的刚性需求。
从商业价值维度看,ReMe 的护城河在于其渐进式混合检索架构:BM25 + wikilink 关系图作为默认检索路径,向量检索作为可选增强,并通过 RRF 融合。这一设计在工程上极其务实——它允许企业以极低的算力成本(无需启动 embedding 模型)获得可用的记忆检索能力,同时保留了向语义检索平滑升级的路径。对比 Mem0、LangMem 等竞品,ReMe 的差异化在于:它不只是一个检索库,而是一个完整的"记忆生命周期管理"系统,涵盖捕获(session)、蒸馏(daily)、固化(digest)、关联(wikilink)、主动推送(proactive)五个环节。
企业级落地场景已较为清晰:编码 Agent 的跨会话经验沉淀(Claude Code 集成)、个人知识管理系统的语义层(LLM Wiki)、金融/科研领域的定时信息蒸馏(Auto Fin、Daily Paper 插件)。特别是 ACL 2026 论文背书和 arXiv 技术报告,为技术选型提供了学术可信度。
2. 核心架构设计与技术深度
ReMe 的架构本质是一个"文件系统为源、索引为影"的双层存储系统,其模块拆解遵循严格的单向数据流:
capture → index → consolidate → recall │ │ │ │ session/ metadata/ daily/ digest/ (源) (可重建) (短期) (长期)
在 README 中展示上榜徽章
[](https://github.fq6825.top/repo/agentscope-ai/ReMe)