← Back to Skills Marketplace
zyuanjia

Long Memory

by ZYuanJia · GitHub ↗ · v7.2.0 · MIT-0
cross-platform ⚠ suspicious
88
Downloads
0
Stars
1
Active Installs
3
Versions
Install in OpenClaw
/install long-memory
Description
全量对话记忆系统，自动保存每次对话的完整内容（用户消息、助手回复、思考过程）并支持按语义检索任意历史对话。当用户提到"记下来""别忘了""回忆一下""之前聊过""什么时候说的"、要求回顾历史对话、或需要跨 session 记忆连续性时触发。也适用于 session 结束时的自动记忆归档。不适用于简单的短期记事（用...
README (SKILL.md)

Long Memory

文件系统 = 无限外存，上下文窗口 = 有限内存。所有对话全量落盘，按需检索，永不遗忘。

企业级特性

  • 并发安全：文件锁机制，多进程写入不冲突
  • 三级搜索引擎：关键词 → TF-IDF + 近义词 → Embedding 向量（零依赖到深度语义）
  • SQLite 引擎：结构化存储，查询比纯文件快 100 倍
  • 自记忆引擎：系统记住自己的版本演变和决策历史
  • 完整性校验：SHA256 校验和，自动检测修复损坏文件
  • 自动备份：Git commit + push，支持定时自动执行
  • 质量评分：0-100 分量化评估（7 条规则）
  • 遗忘曲线：艾宾浩斯遗忘曲线检索权重
  • 情感分析：对话情绪变化追踪
  • 人物图谱：自动提取人脉关系
  • 智能推荐：基于当前对话推荐相关历史
  • 隐私加密：AES-256 加密敏感记忆
  • 结构化模板：会议纪要/项目决策/学习笔记/周报
  • 定时调度：内置蒸馏/备份/清理调度器
  • 多用户隔离：独立记忆空间
  • REST API：7 个 HTTP 接口，支持外部集成
  • HTML 报告：暗色主题可视化分析面板
  • 操作审计：JSONL 完整操作日志
  • 配置管理：JSON 配置 + 版本自动迁移
  • 性能基准：365 天数据模拟，多维度测量
  • 统一 CLI：35 个子命令
  • 37 个测试：pytest 全通过
  • 零核心依赖：搜索引擎、SQLite 全部 Python 标准库

CLI 速查

所有命令支持 --memory-dir 自定义路径，多数命令支持 --json JSON 输出。

📝 记录

cli.py archive     --session webchat --topic "话题"     # 归档对话
cli.py capture     [-s \x3Cid>]                            # 捕获 session 历史
cli.py import      \x3Cfile>                               # 批量导入 MD/JSON/文本
cli.py template    list                                 # 列出模板
cli.py template    create meeting -f participants "张三" # 创建结构化记录

🔍 搜索（三级引擎）

cli.py search      -q "关键词" --tag 标签 --days 30     # 组合搜索（标签+话题+时间）
cli.py semantic    "怎么优化性能"                        # 语义搜索（TF-IDF+近义词，零依赖，推荐）
cli.py tfidf       "查询" --rebuild                     # TF-IDF 向量搜索
cli.py embedding   search "查询" --top 10               # 深度语义搜索（需 sentence-transformers）
cli.py sqlite      search -q "查询"                     # SQLite 高性能查询
cli.py recommend   "当前对话内容"                        # 智能推荐相关历史
cli.py timeline    [-t 话题] [-v]                       # 话题时间线

📊 分析

cli.py stats       [-v]                                 # 统计
cli.py graph       [-v]                                 # 记忆关联图谱
cli.py persons     [-v]                                 # 人物关系
cli.py emotion     [-v]                                 # 情感分析
cli.py health                                           # 健康仪表盘
cli.py report      -o report.html                       # HTML 可视化报告

✅ 质量

cli.py quality                                          # 质量评分（0-100）
cli.py contradictions                                  # 矛盾检测
cli.py integrity   [--fix]                              # 完整性校验
cli.py forgetting                                       # 遗忘曲线

🔧 维护

cli.py summary                                          # 每日摘要
cli.py distill                                          # 周蒸馏
cli.py tags       [--execute]                           # 标签优化
cli.py clean       --days 90 [--execute]                # 清理旧记忆
cli.py index       [--rebuild]                          # 索引管理
cli.py backup      [--status|--setup]                   # Git 备份
cli.py scheduler   [--list|--run|--force]               # 定时任务
cli.py self        [--summary]                          # 自记忆引擎
cli.py config      [--migrate]                          # 配置验证与迁移
cli.py benchmark                                         # 性能基准测试

🔒 安全与审计

cli.py privacy     encrypt "内容" -p 密码               # AES-256 加密
cli.py privacy     decrypt "密文" -p 密码               # 解密
cli.py privacy     scan -f \x3Cfile>                       # 敏感内容扫描
cli.py users       list                                 # 多用户管理
cli.py users       switch \x3Cname>                        # 切换用户
cli.py export      -f json -o backup.json               # 导出 JSON/MD/TXT
cli.py log                                              # 操作审计日志

🌐 API

cli.py api         --port 8765                          # 启动 REST API 服务
# GET /api/health  /api/stats  /api/search?q=  /api/topics  /api/tags  /api/conversations
# POST /api/conversations

测试

cd long-memory && python3 -m pytest tests/ -v  # 37 passed
Usage Guidance
This skill appears to implement a full conversation archival system and mostly does what it says, but review and mitigate before installing: - Inspect auto_backup settings: auto_backup.enabled defaults to false, but if enabled and a git remote is configured, the skill will commit and push entire memory to that remote using your local git credentials — ensure no sensitive data will be pushed and no remote is configured you don't trust. - Confirm git presence and credential helpers: the code calls the git CLI via subprocess; if your system has stored git credentials they may be used automatically. - Restrict where memories are stored: set memory_dir to an isolated directory (not your home or system-wide folders) and review existing files under ~/.openclaw/sessions if you don't want session capture. - Consider disabling or firewalling the API: the HTTP server binds to 127.0.0.1 (local only) but sets Access-Control-Allow-Origin: *. If other local apps/webviews could call it, sensitive data might be accessible; remove permissive CORS or avoid running the API. - Audit capture behavior: capture_session will read OpenClaw session JSON files and archive assistant 'thoughts' — if you consider that private, disable automatic capture/cron jobs. - Run in a sandbox/container first: because the bundle contains many executable scripts that operate on disk and network, test it in an isolated environment before granting access to real conversations. If you want, I can point out the exact lines in auto_backup.py, api_server.py, and capture_session.py that are most relevant and show recommended configuration changes (e.g., disable backup push, remove wildcard CORS, change memory_dir).
Capability Assessment
ℹ Purpose & Capability
Name/description (full conversation memory, semantic search, encryption, backups) is implemented by many included Python scripts. This aligns with the stated purpose. However the package uses the git CLI (subprocess calls) for automatic backups and optionally reads OpenClaw session files under ~/.openclaw/sessions — these runtime dependencies and file accesses are not declared in the skill metadata (it says no required binaries/env).
⚠ Instruction Scope
Runtime instructions and scripts will read and persist full conversation content (including assistant messages and derived metadata), capture OpenClaw session JSON files, run integrity checks, and expose a local REST API. The REST API sets Access-Control-Allow-Origin: * (permitting cross-origin access from other local apps) and there is a backup path that will git commit & push memory data to a remote repository if configured. These behaviors are coherent with archival purpose but expand scope to file-system access, local session harvesting, network exposure, and potential remote transmission of private content.
ℹ Install Mechanism
No install spec is provided (files are bundled). That avoids remote downloads, which is lower risk, but the bundle includes many executable scripts that will run locally. The code executes subprocess git commands (requires 'git' binary) and optionally can use external Python packages (e.g., sentence-transformers for embeddings) even though SKILL metadata claims 'zero core dependencies' — this is a mismatch to be aware of.
⚠ Credentials
The skill declares no required environment variables or credentials, but it will use local git configuration / credential helpers when auto_backup pushes to a remote repo — effectively leveraging existing credentials to transmit archived memories. It also reads ~/.openclaw/sessions (agent session history) and writes to ~/.openclaw/workspace/memory by default. Requesting no explicit credentials while performing actions that can use or expose credentials (git push) is a proportionality concern.
ℹ Persistence & Privilege
The skill does not claim 'always: true' and does not modify other skills, but it writes persistent data to disk, can run a local HTTP server (binds to 127.0.0.1:8765), schedule periodic tasks (scheduler), and perform git push operations. Autonomous invocation is allowed by default; combined with the backup/push and API, this increases potential blast radius if the skill is invoked without careful configuration.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install long-memory
  3. After installation, invoke the skill by name or use /long-memory
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v7.2.0
审计修复: SKILL.md 补全35个命令, .skillignore 排除非skill文件
v7.1.0
v7.1: 自记忆引擎 — 系统记住自己的每一次迭代
v7.0.0
- 增加企业级全量对话记忆系统，支持自动保存、检索和归档全部历史对话内容 - 引入并发安全、内存索引+SQLite+TF-IDF+语义搜索、SHA256完整性校验、自动备份等核心特性 - 新增情感分析、人物图谱、遗忘曲线等多维度数据分析与推荐 - 支持多用户隔离、敏感记忆AES-256加密及多格式记忆导入导出 - 全面强化 CLI，提供 25 个子命令，一站式对话记忆管理 - 完整测试覆盖，内置 37 个 pytest 用例
Metadata
Slug long-memory
Version 7.2.0
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 3
Frequently Asked Questions

What is Long Memory?

全量对话记忆系统，自动保存每次对话的完整内容（用户消息、助手回复、思考过程）并支持按语义检索任意历史对话。当用户提到"记下来""别忘了""回忆一下""之前聊过""什么时候说的"、要求回顾历史对话、或需要跨 session 记忆连续性时触发。也适用于 session 结束时的自动记忆归档。不适用于简单的短期记事（用... It is an AI Agent Skill for Claude Code / OpenClaw, with 88 downloads so far.

How do I install Long Memory?

Run "/install long-memory" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Long Memory free?

Yes, Long Memory is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Long Memory support?

Long Memory is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Long Memory?

It is built and maintained by ZYuanJia (@zyuanjia); the current version is v7.2.0.

More Skills
self-improving agent
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⬇ 463283 · by pskoett
Skill Vetter
Security-first skill vetting for AI agents. Use before installing any skill from ClawdHub, GitHub, or other sources. Checks for red flags, permission scope, and suspicious patterns.
⬇ 259410 · by spclaudehome
Polymarket
Query Polymarket prediction markets. Check odds, find trending markets, search events, track price movements.
⬇ 232503 · by joelchance
Self-Improving + Proactive Agent
Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...
⬇ 200423 · by ivangdavila
Github
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
⬇ 191115 · by steipete
ontology
Typed knowledge graph for structured agent memory and composable skills. Use when creating/querying entities (Person, Project, Task, Event, Document), linkin...
⬇ 190156 · by oswalpalash

💬 Comments