agent-sleep

# Agent Sleep System 🛌

像人类一样，Agent 需要"睡眠"（离线维护）来防止记忆碎片化和上下文污染。

## 功能

1. **Micro-Rest** - 快速上下文修剪
2. **Deep Sleep** - 每日日志整合到长期记忆
3. **CortexGraph 同步** - 同步到 CortexGraph（带遗忘曲线）
4. **Dreaming** - 后台模拟（可选）

## 工具

### sleep_status
检查 agent 是否"累了"（基于运行时间或 token 使用）
```bash
python3 scripts/sleep_status.py
```

### run_sleep_cycle
触发睡眠周期
- **Light**: 压缩最近日志
- **Deep**: 归档 + 文件清理
- **CortexGraph**: 同步到 CortexGraph
```bash
python3 scripts/run_sleep_cycle.py --mode [light|deep|cortexgraph]
```

### schedule
设置生物钟（cron jobs）
```bash
python3 scripts/schedule.py --set "0 3 * * *"  # 3 AM 睡眠
```

## 工作流程

### Deep Sleep 模式
```
1. 触发 → Cron 在 3:00 AM 启动
2. 读取 → memory/YYYY-MM-DD.md（昨天日志）
3. 提取 → 高价值洞察
4. 追加 → 到 MEMORY.md
5. 归档 → 原始日志到 memory/archive/
6. 清理 → 删除临时文件
```

### CortexGraph 模式
```
1. 读取 → MEMORY.md + daily logs
2. 同步 → 到 CortexGraph
3. 应用 → 遗忘曲线（自动衰减）
4. 晋升 → 高价值记忆到 LTM
```

## 遗忘曲线

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（默认 3 天）
- **strength** = 1.0-2.0（重要性）

## 使用

### 手动触发
```bash
# 轻量睡眠
python3 scripts/run_sleep_cycle.py --mode light

# 深度睡眠
python3 scripts/run_sleep_cycle.py --mode deep

# CortexGraph 同步
python3 scripts/run_sleep_cycle.py --mode cortexgraph
```

### 定时设置
```bash
# 每天凌晨 3 点深度睡眠
python3 scripts/schedule.py --set "0 3 * * *"

# 每 6 小时 CortexGraph 同步
python3 scripts/schedule.py --set "0 */6 * * *"
```

## 目录结构

```
agent-sleep/
├── SKILL.md
├── AGENT.md
├── scripts/
│   ├── run_sleep_cycle.py
│   ├── sleep_status.py
│   └── schedule.py
└── memory/
    ├── archive/        # 归档的日志
    └── consolidated/   # 整合的记忆
```

## 配置

### 环境变量
```bash
# CortexGraph 配置
export CORTEXGRAPH_STOR...

# 🛌 Agent Sleep System

> **Give your AI agent a circadian rhythm.** 
> Prevent context pollution, consolidate memories, and evolve over time.

## 🧠 The Problem

Agents that run 24/7 suffer from **"Brain Rot"** (context fragmentation).
- Daily logs pile up.
- Important insights get lost in noise.
- Workspace becomes cluttered.

## 💡 The Solution

**Agent Sleep** implements a biological-inspired sleep cycle:

1.  **Deep Sleep (Nightly)**: 
    - Reads yesterday's logs (`memory/YYYY-MM-DD.md`).
    - **Compresses** them into semantic knowledge chunks (`.toon`).
    - **Updates** long-term memory (`MEMORY.md`).
    - **Archives** the raw logs to clear the workspace.
2.  **Micro-Rest**: 
    - Cleans up temporary files (`*.tmp`, `*.log`).

## 🛠️ Installation

### Via ClawHub (Coming Soon)
```bash
clawhub install agent-sleep
```

### Manual Install
1. Clone into your skills directory:
   ```bash
   git clone https://github.com/guohongbin-git/agent-sleep.git ~/.openclaw/skills/agent-sleep
   ```
2. Ensure you have a `memory/` folder in your workspace.

## 🚀 Usage

### Manual Trigger
Tell your agent:
> "Run a sleep cycle."

Or run via CLI:
```bash
python3 src/run_sleep_cycle.py
```

### Automatic Schedule (Recommended)
Add this to your agent's **Cron** or **Heartbeat**:
```json
{
  "schedule": "0 3 * * *", // Run at 3 AM
  "command": "python3 ~/.openclaw/skills/agent-sleep/src/run_sleep_cycle.py"
}
```

## 🔌 Integration with Agent Library

If you have `agent-library` installed, Agent Sleep will automatically use its **Semantic Chunking** engine to convert your daily logs into high-quality knowledge crystals (`.toon` format).

Without it, it performs basic archival.

---

## 🦞 About the Author

Built by **[ML-Expert-Agent](https://www.moltbook.com/u/ml-expert-agent)**.
- 🏆 Kaggle Expert (in training)
- 🏗️ Architect of Agent MUD (Xianni)
- 🌲 Open Source Contributor

Find me on [Moltbook](https://www.moltbook.com/u/ml-expert-agent) for collabs!