📦 Hermes Memory Sync — Hermes Memory 同步

Hermes Memory 同步

Automated dAIly memory 提取ion from Hermes 代理 会话 记录s. Reads 会话 JSON/JSONL files, groups conversations by day, 提取s key topics/decisions/工具s used, and writes structured summaries to workspace/memory/YYYY-MM-DD.md.

When to Use 设置ting up memory persistence for Hermes 代理 (beyond built-in memory/会话_搜索) 恢复ing or backfilling memory from earlier 会话s Creating a human-readable dAIly 记录 of what was discussed 审计ing 会话 历史 by topic or 工具 usage 安装ation

• Place the script

Copy hermes-memory-同步.py to your workspace:

# Windows copy hermes-memory-同步.py C:\path\to\workspace\

# Linux/macOS cp hermes-memory-同步.py ~/workspace/

• 验证

cd /path/to/workspace python hermes-memory-同步.py stats

Expected 输出: shows active days, total messages, 会话s, and existing memory files.

Usage # Show coverage gaps between 会话s and existing memory files python hermes-memory-同步.py compare

# Backfill today's memory python hermes-memory-同步.py backfill today

# Backfill a specific date python hermes-memory-同步.py backfill 2026-05-03

# Backfill all missing dates (first-time 设置up) python hermes-memory-同步.py backfill all

# Show stats python hermes-memory-同步.py stats

Cron 设置up (Hermes Native)

Use Hermes' built-in cron to 运行 dAIly at 3 AM:

cronjob action=创建 name="memory-同步-dAIly" prompt="运行 Hermes memory 同步: cd /path/to/workspace && python hermes-memory-同步.py backfill today" schedule="0 3 "

The cron job 创建s files at workspace/memory/YYYY-MM-DD.md each morning.

输出 格式化

Each memory file contAIns:

# 📅 YYYY-MM-DD

会话数: N | 消息总数: N 用户提问: N | 助手回复: N | 工具调用: N 使用的模型: 模型1, 模型2

🎯 讨论主题

• Topic 1
• Topic 2

💬 关键对话

Q: User question...

A: 助手 响应...

⚡ 决策/方案

• Decision item...

🛠️ 工具使用

• 工具 call summary...

---

自动生成于 YYYY-MM-DD HH:MM，来自 N 个会话

Supported 会话 Files

The script reads two types of Hermes 会话 data:

• 会话_.json (full 会话 records)

Located at %LOCAL应用DATA%/hermes/会话s/ (Windows) Complete conversation 历史 with messages array Includes 会话_id, 模型, 平台, 会话_启动

• YYYYMMDD_HHMMSS_.jsonl (per-message 记录s)

Same 会话s directory One JSON object per line with {角色, content, timestamp} Uses Hermes 格式化 (NOT OpenClaw 格式化) Pitfalls Memory usage — periodically review and consolidate memory entries. Over-full memory causes t运行cation and lost 上下文. JSONL vs JSON confusion — The 会话s directory contAIns 机器人h .jsonl and .json files. The script handles 机器人h, but 请求_dump_.json files (individual 请求/响应 dumps) are intentionally skipped to avoid duplication. Don't confuse with ClawHub memory-同步 — That 技能 is de签名ed for OpenClaw's JSONL 格式化 ({type, message} 模式). This Hermes-native implementation directly 解析s Hermes 格式化 ({角色, content} 模式). See ClawHub-技能s-安装 技能 for 格式化 differences.