---
name: memory-compaction-b36bf9f18bd5
description: "OpenClaw \u00d7 Claude Code \u7cbe\u534e\u6574\u5408 \u2014 \u8bb0\u5fc6\u7cfb\u7edf\u3001\u63d0\u793a\u8bcd\u67b6\u6784\u3001\u7194\u65ad\u673a\u5236"
---

# memory-compaction

## 用途

定期清理 LanceDB 记忆碎片，防止只进不出导致检索质量下降。

来源：Claude Code 的 autoDream 系统思路，针对 OpenClaw + LanceDB 重新实现。

---

## 安装方式

复制到 `~/.openclaw/workspace/skills/memory-compaction/SKILL.md`，
然后发给你的 OpenClaw：

```
帮我创建 memory-compaction skill 的执行脚本，并注册每周日凌晨 3 点的 cron job。
使用现有的 SiliconFlow bge-m3 做相似度计算，不引入新的 embedding provider。
```

---

## 执行逻辑

每周日 03:00 自动触发，按顺序执行：

**第一步：备份**
```
~/.openclaw/memory/lancedb-pro/backups/YYYY-MM-DD/
保留最近 4 份，更早的自动删除
```

**第二步：删除低价值记忆**
- 条件：importance < 0.3 且距上次检索超过 14 天
- 遇到任何错误：立即停止，发告警，不继续

**第三步：合并相似碎片**
- 条件：同主题下向量相似度 ≥ 0.85
- 合并策略：保留 importance 最高的那条的 L0 摘要，其余删除

**第四步：输出报告**
```
🧠 Memory Compaction 报告
删除：N 条
合并：M 簇
当前总数：X 条
耗时：Xs
```

---

## 首次使用建议

正式执行前先做 dry-run：

```
触发一次 memory-compaction 的 dry-run，只报告会删除/合并哪些，不实际执行
```

确认没有误删风险后再正式执行。

---

## 常见问题

**备份目录不存在导致失败：**
```bash
mkdir -p ~/.openclaw/memory/lancedb-pro/backups
```

**LanceDB SQL 不支持 list 类型：**
执行 update() 时需排除 vector 字段，这是 LanceDB 的已知限制。

---

## 与 compact-guardian 的关系

| skill | 解决的问题 | 触发方式 |
|---|---|---|
| compact-guardian | auto-compact 失败熔断 | 实时，每次 compact 失败时 |
| memory-compaction | LanceDB 碎片堆积 | 定时，每周日 03:00 |

两者解决不同层面的问题，都需要安装。


---
Source attribution: Ghenghis
Original project: https://github.com/Ghenghis/openclaw-claude-code-integration/blob/c96bed8f902f77d8ef75632000539fb17789a25a/skills/memory-compaction/SKILL.md
License: MIT
License information: https://opensource.org/license/mit
Original upstream instructions; downloaded from Mega AI Hub.

## Upstream license notice

MIT License

Copyright (c) 2026 ybbms

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.



# Mega AI Hub resource metadata

Creator / publisher: Ghenghis
Original project: https://github.com/Ghenghis/openclaw-claude-code-integration/blob/c96bed8f902f77d8ef75632000539fb17789a25a/skills/memory-compaction/SKILL.md
Attribution basis: Source repository record

Fields of work: Development & Engineering
Roles: Data Engineer, DevOps / SRE, ML / AI Engineer
Specialties: Backend
Tools: Tool not specified
Country applicability: Country not specified
Classification basis: Source catalog labels
Instruction SHA-256: 01646a30f81a23c923fe2f6252f03ed72c7f23f29e1a7bfca7d2fc2b32feb550