Skip to content

susu5/elite-longterm-memory

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🧠 Elite Longterm Memory

Ultimate AI agent memory system — WAL protocol + vector search + git-notes + cloud backup.


中文描述

Elite Longterm Memory 是一套面向 AI Agent 的终极长期记忆系统,融合 6 种成熟方案构建出高可靠性架构:

  • HOT RAM(会话级):SESSION-STATE.md,在上下文压缩后仍然存活
  • WARM STORE(语义检索):LanceDB 向量数据库,支持语义搜索
  • COLD STORE(永久决策):Git-Notes 知识图谱
  • MEMORY.md:人工可读的长期记忆策展
  • SuperMemory API:云端备份(可选)

核心能力: 永不丢失上下文、永不遗忘决策、永不重复犯错。

使用方式

  1. 安装依赖:npm install
  2. 配置环境变量:OPENAI_API_KEY
  3. 安装插件:memory-lancedb
  4. 在 AI Agent(Cursor / Claude / ChatGPT / Copilot)中加载此 Skill
  5. Agent 将自动在对话中读写记忆,跨会话保持上下文连续性

触发场景

  • 需要跨对话保持项目上下文
  • 需要语义搜索历史决策
  • 需要 WAL 协议保证记忆写入可靠性

English Description

Elite Longterm Memory is the ultimate memory system for AI agents, combining 6 proven approaches into one bulletproof architecture:

  • HOT RAM (session-level): SESSION-STATE.md survives context compaction
  • WARM STORE (semantic search): LanceDB vector database
  • COLD STORE (permanent decisions): Git-Notes knowledge graph
  • MEMORY.md: Curated human-readable long-term memory
  • SuperMemory API: Cloud backup (optional)

Core promise: Never lose context. Never forget decisions. Never repeat mistakes.

Usage

  1. Install dependencies: npm install
  2. Set environment variable: OPENAI_API_KEY
  3. Install plugin: memory-lancedb
  4. Load this Skill in your AI Agent (Cursor / Claude / ChatGPT / Copilot)
  5. The Agent will automatically read/write memory across sessions

Tech Stack

  • TypeScript / Node.js
  • LanceDB (vector search)
  • Git-Notes (knowledge graph)
  • WAL protocol (write-ahead logging)

License

MIT

About

🧠 Ultimate AI agent memory system - WAL protocol + vector search + git-notes + cloud backup. AI Agent 终极长期记忆系统 | License: CC-BY 4.0

Topics

Resources

License

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors