lobster-memory（龙虾记忆）

龙虾记忆大师 (Lobster Memory Master)

真正的长期记忆管理方案 Created by: 天道桐哥 & AI龙虾元龙 🦞

核心功能

1. 自动记忆维护 (Auto Memory Maintenance)

当 Context 超过 60% 时自动触发：

• 归档到 memory/YYYY-MM-DD.md
• 更新 SESSION-STATE.md
• 清理 Working Buffer

2. 每日学习日志 (Daily Learning Log)

自动记录：

• 技能安装/学习记录
• 项目进展
• 关键决策
• 用户偏好

3. 心跳协议 (Heartbeat Protocol)

定期自检：

• Context 使用率检查
• 记忆文件归档
• 主动行为建议

4. 工作缓冲区 (Working Buffer)

危险区管理：

• Context > 60% 时进入
• 记录关键对话片段
• 防止重要信息丢失

快速开始

初始化记忆系统

# 创建必要文件
touch MEMORY.md
mkdir -p memory
touch memory/working-buffer.md
touch SESSION-STATE.md
touch HEARTBEAT.md

写入 MEMORY.md

# MEMORY.md - Long-Term Memory

## About [User Name]
- Name: [User Name]
- Style: [User Style]
- Project: [Current Project]

## Key Decisions
- [Important decision 1]
- [Important decision 2]

## Active Projects
- [Project 1] - [Status]
- [Project 2] - [Status]

设置定时任务

# Daily Memory Freshener
cron:
  - name: "Memory Freshener"
    schedule: "0 2 * * *"  # 每天凌晨2点
    action: archive_memory

# Daily Learning
cron:
  - name: "Daily Skill Discovery"
    schedule: "0 10 * * *"  # 每天上午10点
    action: learn_new_skills

文件结构

workspace/
├── MEMORY.md                 # 长期记忆主文件
├── SESSION-STATE.md          # 当前会话状态
├── HEARTBEAT.md             # 心跳检查清单
├── AGENTS.md                # 代理操作规则
├── memory/
│   ├── working-buffer.md    # 工作缓冲区
│   ├── YYYY-MM-DD.md        # 每日归档
│   └── learning-log.md      # 学习日志
└── ...

使用场景

场景1：用户问起之前的事

触发条件： User asks about prior work, decisions, dates, people, preferences

动作：

1. Search MEMORY.md + memory/*.md
2. Pull only needed lines with memory_get
3. Answer with citation: Source: <path#line>

场景2：Context 超过 60%

触发条件： Context usage > 60%

动作：

1. Enter Working Buffer protocol
2. Archive to memory/YYYY-MM-DD.md
3. Update SESSION-STATE.md
4. Clear working buffer

场景3：设置自主学习

触发条件： User wants autonomous daily learning

动作：

1. Create AGENTS.md with learning protocols
2. Set up cron jobs for daily tasks
3. Create learning-log.md with rotation schedule

最佳实践

记忆搜索流程

1. memory_search(query) → 找相关片段
2. memory_get(path, from, lines) → 拉取具体内容
3. Answer with citation

归档流程

1. Check context %
2. If > 60%: Archive working buffer
3. Update SESSION-STATE.md
4. Clear old entries

学习日志流程

1. Record daily activities
2. Distill to MEMORY.md weekly
3. Rotate old logs monthly

示例代码

自动归档脚本

// scripts/archive-memory.js
const fs = require('fs');
const path = require('path');

function archiveMemory() {
  const today = new Date().toISOString().split('T')[0];
  const bufferFile = 'memory/working-buffer.md';
  const archiveFile = `memory/${today}.md`;
  
  // Read working buffer
  const buffer = fs.readFileSync(bufferFile, 'utf8');
  
  // Append to archive
  fs.appendFileSync(archiveFile, buffer);
  
  // Clear buffer
  fs.writeFileSync(bufferFile, '# Working Buffer\
\
**Status:** CLEARED\
');
  
  console.log(`✅ Archived to ${archiveFile}`);
}

archiveMemory();

备注

本技能由天道桐哥 & AI龙虾元龙共同完成 🦞

• 天道桐哥：Human Creator, Product Vision
• AI龙虾元龙：AI Creator, Implementation

Created: 2026-03-22 Version: 1.0.0 License: MIT