memory-sync-cn

记忆同步系统 - 自动同步 MEMORY.md 与 CortexGraph,支持遗忘曲线和智能检索

Safety Notice

This listing is from the official public ClawHub registry. Review SKILL.md and referenced scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "memory-sync-cn" with this command: npx skills add guohongbin-git/memory-sync-cn

记忆同步系统

自动同步 OpenClaw 记忆系统与 CortexGraph(带遗忘曲线)。

架构

MEMORY.md (长期记忆)
     ↕
CortexGraph (智能层)
     ↕
memory/YYYY-MM-DD.md (每日日志)

同步脚本

1. 导入 MEMORY.md 到 CortexGraph

# 运行同步
./scripts/sync-memory.sh

2. 导入每日日志

# 同步今天的日志
./scripts/sync-daily.sh

# 同步指定日期
./scripts/sync-daily.sh 2026-02-18

3. 晋升高价值记忆

# 查看晋升候选
mcporter call cortexgraph.promote_memory auto_detect=true dry_run=true

# 执行晋升
mcporter call cortexgraph.promote_memory auto_detect=true

CortexGraph 工具

核心操作

# 保存记忆
mcporter call cortexgraph.save_memory content="..." tags='["tag1"]' entities='["Entity1"]'

# 搜索记忆
mcporter call cortexgraph.search_memory query="关键词" top_k=5

# 强化记忆(减缓衰减)
mcporter call cortexgraph.touch_memory memory_id="UUID" boost_strength=true

# 查看知识图谱
mcporter call cortexgraph.read_graph limit=10

智能分析

# 分析是否值得记住
mcporter call cortexgraph.analyze_message message="用户说的话"

# 分析是否需要搜索
mcporter call cortexgraph.analyze_for_recall message="用户说的话"

# 自动回忆
mcporter call cortexgraph.auto_recall_process_message message="用户说的话"

维护

# 垃圾回收(删除低分记忆)
mcporter call cortexgraph.gc dry_run=true

# 合并相似记忆
mcporter call cortexgraph.consolidate_memories auto_detect=true mode=preview

# 生成嵌入向量
mcporter call cortexgraph.backfill_embeddings dry_run=true

遗忘曲线算法

CortexGraph 使用 Ebbinghaus 遗忘曲线:

score = (use_count)^β × e^(-λ × Δt) × strength
  • β = 0.6(使用频率权重)
  • λ = ln(2) / half_life(衰减常数,默认 3 天)
  • strength = 1.0-2.0(重要性)

记忆会自然衰减,除非:

  • 被引用(touch_memory)
  • 被搜索到(auto_recall)
  • 晋升到长期记忆(promote_memory)

配置

编辑 ~/.config/cortexgraph/.env

# 存储路径
CORTEXGRAPH_STORAGE_PATH=~/.config/cortexgraph/jsonl

# 衰减参数
CORTEXGRAPH_DECAY_MODEL=power_law
CORTEXGRAPH_PL_HALFLIFE_DAYS=3.0

# 阈值
CORTEXGRAPH_FORGET_THRESHOLD=0.05
CORTEXGRAPH_PROMOTE_THRESHOLD=0.65

# 长期记忆(Obsidian)
LTM_VAULT_PATH=~/Documents/Obsidian/Vault

使用场景

1. 对话中自动记住

# 分析消息
mcporter call cortexgraph.analyze_message message="用户说..."
# 如果 should_save=true,则保存

2. 对话中自动回忆

# 分析消息
mcporter call cortexgraph.analyze_for_recall message="用户问..."
# 如果 should_search=true,则搜索

3. 定期维护

# 每周运行
mcporter call cortexgraph.gc
mcporter call cortexgraph.consolidate_memories auto_detect=true mode=apply

与 OpenClaw 集成

CortexGraph 已通过 mcporter 配置:

// ~/.openclaw/workspace/config/mcporter.json
{
  "cortexgraph": {
    "command": "/home/ghb/.local/bin/cortexgraph",
    "description": "Temporal memory system for AI with Ebbinghaus forgetting curve"
  }
}

最佳实践

  1. 重要信息 strength=1.5-2.0(用户偏好、关键决策)
  2. 普通信息 strength=1.0(日常对话)
  3. 临时信息 strength=0.5(上下文相关)
  4. 每周维护:GC + 合并 + 晋升
  5. 每月检查:调整衰减参数

版本: 1.0.0 作者: 赚钱小能手

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

Unified Memory

Provides a layered, atomic, and performant memory management system with structured recall, vector search, smart deduplication, compression, and lifecycle ma...

Registry SourceRecently Updated
1370Profile unavailable
General

Unified Memory V5

Provides a layered memory management system with atomic transactions, smart deduplication, vector search optimization, and lifecycle management for consisten...

Registry SourceRecently Updated
1190Profile unavailable
General

vemem — visual entity memory

Visual entity memory — remember faces, objects, and places across sessions with persistent identity. Use when the user asks who is in an image, when you need...

Registry SourceRecently Updated
720Profile unavailable
General

Pieces Long-Term Memory (MCP)

Connect OpenClaw to Pieces via MCP-only (no SSE) and use Pieces as external long-term memory. Use this when the human runs PiecesOS with LTM enabled on anoth...

Registry SourceRecently Updated
2492Profile unavailable