← Back to Skills Marketplace
Hongnao Memory V1.0.0
by
tangfengxm
· GitHub ↗
· v1.0.0
· MIT-0
68
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install hongnao-memory-v1-0-0
Description
为 OpenClaw 提供长期记忆管理，支持跨会话持久化、智能检索和用户偏好学习，提升记忆存储与检索效率。
README (SKILL.md)
弘脑记忆系统 (HongNao Memory OS)
为 OpenClaw 打造的长期记忆操作系统，实现跨 Session 记忆持久化、智能检索和用户偏好学习。
快速开始
安装
# 自动安装 (推荐)
python3 install_hongnao.py
# 或手动安装
pip3 install -r requirements.txt
基本使用
from memory_api import HongNaoMemorySystem, MemorySystemConfig
# 初始化
config = MemorySystemConfig()
memory_system = HongNaoMemorySystem(config)
# 添加记忆
memory_system.add_memories_from_text("用户喜欢简洁的沟通风格", source="chat")
# 检索记忆
result = memory_system.retrieve_memories("用户偏好")
print(result['context'])
核心功能
- 4 层记忆架构: 抽取 → 巩固 → 检索 → 更新
- OpenClaw 集成: Session 自动同步到记忆系统
- 用户偏好学习: 自动识别用户习惯和偏好
- 混合检索: 向量 + 关键词，召回率>90%
- 记忆压缩: Token 节省~45%
性能指标
| 指标 | 目标 | 实测 |
|---|---|---|
| 添加延迟 | \x3C10ms | 0.03ms ✅ |
| 检索延迟 | \x3C100ms | 0.27ms ✅ |
| 召回率 | >90% | 100.0% ✅ |
| Token 节省 | >40% | ~45% ✅ |
配置
编辑 hongnao_memory_config.ini:
[memory]
db_path = ~/.openclaw/workspace/hongnao_memory/hongnao.db
vector_store_path = ~/.openclaw/workspace/hongnao_memory/chroma_db
[preference_learning]
enable_preference_learning = true
confidence_threshold = 0.6
文档
- 快速入门:
快速入门.md - 集成指南:
集成指南.md - 发布说明:
RELEASE_NOTES.md
测试
# 核心模块测试
python3 test_core_modules.py
# 性能基准测试
python3 quick_benchmark.py
系统要求
- Python >= 3.8
- OpenClaw >= 0.1.0
- 支持平台：Linux, macOS, Windows
依赖
- chromadb >= 0.4.0
- numpy >= 1.20.0
- sentence-transformers >= 2.2.0
许可证
MIT License
联系方式
- 作者：燧弘华创
- 邮箱：[email protected]
- 主页：https://github.com/hongnao/hongnao_memory
Usage Guidance
This package appears to be a legitimate OpenClaw memory plugin, but it contains multiple red flags from poor engineering rather than clear maliciousness: files referenced in docs or installer are missing, some modules import inconsistent filenames (memory_extraction vs memory_extraction_v3), and docs make unrealistically strong performance claims. The installer will copy code into your ~/.openclaw/workspace and create config files — review the code first. Recommended steps before installing: 1) Inspect openclaw_integration.py and any truncated/omitted files for network calls or endpoints; 2) Run the installer in an isolated/sandbox environment or a disposable OpenClaw workspace; 3) Fix or validate the import/name mismatches and missing files (or ask the publisher for a clean release); 4) Run the included tests (in sandbox) to see runtime errors; 5) Only enable autonomous invocation after you vet the code. If you cannot validate the missing/mismatched files or the publisher, treat this as untrusted and avoid installing into a production OpenClaw workspace.
Capability Analysis
Type: OpenClaw Skill
Name: hongnao-memory-v1-0-0
Version: 1.0.0
The skill bundle implements a comprehensive long-term memory system for OpenClaw agents, featuring memory extraction, consolidation, hybrid retrieval (vector + keyword), and a forgetting mechanism. The code is well-structured and uses local SQLite and ChromaDB storage within the designated OpenClaw workspace (~/.openclaw/workspace). Analysis of the installation script (install_hongnao.py), core logic (memory_api.py, storage_layer.py), and agent instructions (SKILL.md) reveals no evidence of data exfiltration, malicious execution, or unauthorized persistence. All functionalities, including user preference learning and session synchronization, are strictly aligned with the stated purpose of providing persistent memory for AI sessions.
Capability Assessment
Purpose & Capability
Name, description and most files align with a long‑term memory plugin for OpenClaw (extraction, consolidation, retrieval, update, preference learning). However the registry/skill metadata claims 'instruction-only' while many code files are bundled — an internal mismatch. Some referenced files in docs (main.py, test_session_integration.py) are missing from the manifest or file list, and several modules import different filenames (memory_consolidation imports memory_extraction while actual extraction file is memory_extraction_v3.py). These inconsistencies are not expected for a production release.
Instruction Scope
SKILL.md and README instruct running install_hongnao.py which copies plugin code into the user's OpenClaw workspace and creates a config file (~/ .openclaw/workspace/hongnao_memory). For a memory plugin, copying files and creating configs is expected, and the code legitimately needs access to session data for syncing. But the instructions reference files/entrypoints that are missing or mismatched, and installer prompts (and sample code) give broad discretion to auto‑sync sessions — inspect openclaw_integration.py before granting access. No explicit instructions were found to exfiltrate secrets, but the installer will write files into the user's home workspace.
Install Mechanism
Installer is a local Python script (install_hongnao.py) that copies bundled files and writes a config and requirements.txt — no remote downloads or URL fetches observed in the provided files. This is lower risk than remote installers, but it will place code into the user's workspace and create configuration files.
Credentials
The skill declares no required environment variables, no credentials, and no special config paths beyond writing its own config under the OpenClaw workspace. That matches the described purpose. (Note: the installer checks/imports modules and suggests installing 'sqlite3' via pip which is incorrect — sqlite3 is stdlib; minor sloppy detail.)
Persistence & Privilege
always:false (good). The installer copies files into the user's ~/.openclaw/workspace (or specified workspace) and creates configuration files — this gives persistent presence within the OpenClaw workspace but does not modify other skills or system settings. Because autonomous invocation is allowed by default, the agent could use the integrated memory features automatically; if you do not trust the code, disable autonomous invocation or review code first.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install hongnao-memory-v1-0-0 - After installation, invoke the skill by name or use
/hongnao-memory-v1-0-0 - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- 首次发布：推出 HongNao 记忆操作系统 1.0.0
- 支持 OpenClaw 集成，实现跨 Session 记忆持久化
- 提供 4 层记忆架构：抽取、巩固、检索、更新
- 实现混合检索（向量 + 关键词），召回率达 100%
- 自动学习并识别用户偏好
- 引入记忆压缩，节省约 45% Token
- 性能优异：添加与检索延迟低于 1ms/1s
- 提供详细配置说明及测试脚本
Metadata
Frequently Asked Questions
What is Hongnao Memory V1.0.0?
为 OpenClaw 提供长期记忆管理，支持跨会话持久化、智能检索和用户偏好学习，提升记忆存储与检索效率。 It is an AI Agent Skill for Claude Code / OpenClaw, with 68 downloads so far.
How do I install Hongnao Memory V1.0.0?
Run "/install hongnao-memory-v1-0-0" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Hongnao Memory V1.0.0 free?
Yes, Hongnao Memory V1.0.0 is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Hongnao Memory V1.0.0 support?
Hongnao Memory V1.0.0 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Hongnao Memory V1.0.0?
It is built and maintained by tangfengxm (@tangfengxm); the current version is v1.0.0.
More Skills