360 Secure SkillHub
360 安全技能中心

Memory Master

这是受Hermes Agent启发的AI记忆管理系统,核心功能含迭代压缩(新内容+旧摘要生成累积摘要)、完整记忆谱系追踪、增量检测仅压缩变更内容,还能提取决策/任务/时间线的结构化摘要,性能提升5.6倍。适配AI Agent记忆积累场景,输入为记忆片段或新增内容,输出为压缩后的结构化记忆及谱系链,缓存机制优化效率。

cp3d1455926-svg
cp3d1455926-svg
热度 35.60
最近更新 2026-05-11 17:42:10
分类:开发依赖:代码执行 / 读写文件 / 内部调用大模型应用场景:数据处理 / 摘要总结 / 自动化脚本

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add openclaw-memory-master-v4
Metadata
name
openclaw-memory-master
version
4.2.0
description
AI Memory System with Iterative Compression & Lineage Tracking - Hermes Agent Inspired
author
小鬼 👻 + Jake
tags
memory, ai, agent, iterative-compression, lineage, hermes, openclaw
repository
{"type":"git","url":"https://github.com/cp3d1455926-svg/memory-master.git"}
bugs
{"url":"https://github.com/cp3d1455926-svg/memory-master/issues"}
homepage
https://cp3d1455926-svg.github.io/openclaw-memory/

OpenClaw Memory Master v4.2.0

AI Memory System with Iterative Compression & Lineage Tracking

Inspired by Hermes Agent's "Compression as Consolidation" methodology.

✨ v4.2.0 New Features

  • 🔥 Iterative Compression: New content + old summary → updated summary (accumulative)
  • 🌳 Lineage Tracking: Complete memory genealogy chain
  • ⚡ Performance: 5.6x faster (45ms average, was 250ms)
  • 📊 Structured Summary: Extract decisions/tasks/timeline
  • 💾 Incremental Detection: Only compress changed content

📊 Performance Metrics

Metric v4.1.0 v4.2.0 Improvement
Average Time 250ms 45ms 5.6x ⚡
P95 Latency 400ms 52ms 7.7x ⚡
Cache Hit Rate 0% 72% +72% 🎯

🚀 Quick Start

# Install from ClawHub
clawhub install openclaw-memory-master@4.2.0

# Or clone from GitHub
git clone https://github.com/cp3d1455926-svg/memory-master.git
cd memory-master

📖 Documentation

🧪 Testing

# Run full test suite
node test-full-suite.js

# All 12 tests passed ✅

📝 Changelog

v4.2.0 (2026-04-11)

New Features:

  • Iterative compression with accumulative summaries
  • Lineage tracking for complete memory history
  • LRU cache for 72% hit rate
  • Parallel processing for batch compression
  • Incremental detection to avoid unnecessary compression

Performance:

  • 5.6x faster compression (250ms → 45ms)
  • 7.7x better P95 latency (400ms → 52ms)
  • 72% cache hit rate

Testing:

  • 12/12 tests passed
  • Full coverage of compression, caching, and lineage

v4.1.0 (2026-04-10)

  • Importance scoring system
  • Emotional dimension tracking
  • Dynamic Top-K retrieval

🤝 Contributing

Contributions welcome! Please read CONTRIBUTING.md first.

📄 License

MIT License © 2026 Jake & 小鬼 👻

🔗 Links


Built with ❤️ by Jake & 小鬼 👻

v4.2.0 - Let AI truly "remember" and "grow"! 🚀