← 返回 Skills 市场
wenshuangl

wenshuangl/agent-mem

作者 wenshuangl · GitHub ↗ · v1.0.3 · MIT-0
cross-platform ⚠ pending
91
总下载
1
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install agent-mem
功能描述
Multi-Agent Memory + Dispatch System. 4-tier memory (HOT/WARM/COLD/ARCHIVE), cross-channel sharing, dispatch loop with auto-learning.
使用说明 (SKILL.md)

AgentMem

Multi-Agent Memory + Dispatch System

Core Capabilities

1. Four-Tier Memory (HOT → WARM → COLD → ARCHIVE)

Memories decay naturally over time instead of being treated equally.

2. Cross-Channel Memory Sharing

Same agent shares memory across different channels (webchat/Feishu/Slack/Telegram).

3. Dispatch + Memory Loop

User request → Intent recognition → Agent dispatch → Execution → Auto-log → Optimize next dispatch

4. 17 Memory Modules

Fact extraction, BM25+vector fusion search, contradiction detection, knowledge graph, forgetting mechanism, active recall, memory feedback, self-review.

Quick Start

pip install -e .

# Write a memory
python -m agent_mem.core.hot_cache write --agent main --channel webchat --text "User prefers concise answers" --importance 7

# Cross-channel query
python -m agent_mem.core.hot_cache query --agent main --limit 5

# Dispatch stats
python -m agent_mem.core.dispatch_logger stats

# Run memory engine
python -m agent_mem.core.engine_v2 --mode daily

Requirements

  • Python 3.10+
  • chromadb (single dependency)
  • Zero external API dependencies, fully local

Links

能力标签
cryptorequires-sensitive-credentials
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install agent-mem
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agent-mem 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.3
- Initial public release of agent-mem version 1.0.3. - Introduced 18 new modules including memory tiers, knowledge graph, active recall, forgetting mechanism, and contradiction detection. - Features a four-tier memory system (HOT/WARM/COLD/ARCHIVE) with natural memory decay. - Enables cross-channel memory sharing and a dispatch loop with auto-learning. - Memory modules cover extraction, search, feedback, self-review, and more—all fully local with minimal dependencies.
v1.0.1
- Initial public release of agent-mem (v1.0.1) - Added four-tier memory system (HOT/WARM/COLD/ARCHIVE) with natural decay - Implemented cross-channel memory sharing for agents - Introduced dispatch and memory loop system with auto-learning and optimization - Provided 17 modular memory features including fact extraction, vector search, contradiction detection, and more - Included quick start scripts, setup files, and essential documentation
v1.0.0
agent-mem 1.0.0 - Initial release of a multi-agent memory and dispatch system. - Features a four-tier memory system (HOT/WARM/COLD/ARCHIVE) with natural memory decay. - Enables cross-channel memory sharing for agents across platforms (webchat, Feishu, Slack, Telegram). - Includes a dispatch and memory loop for automated learning and optimization. - Bundles 17 memory modules (fact extraction, search, contradiction detection, knowledge graph, forgetting, active recall, feedback, self-review, and more). - Designed to be local-first with minimal dependencies (Python 3.10+, chromadb, no external APIs).
元数据
Slug agent-mem
版本 1.0.3
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 3
常见问题

wenshuangl/agent-mem 是什么?

Multi-Agent Memory + Dispatch System. 4-tier memory (HOT/WARM/COLD/ARCHIVE), cross-channel sharing, dispatch loop with auto-learning. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 91 次。

如何安装 wenshuangl/agent-mem?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-mem」即可一键安装,无需额外配置。

wenshuangl/agent-mem 是免费的吗?

是的,wenshuangl/agent-mem 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

wenshuangl/agent-mem 支持哪些平台?

wenshuangl/agent-mem 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 wenshuangl/agent-mem?

由 wenshuangl(@wenshuangl)开发并维护,当前版本 v1.0.3。

💬 留言讨论