DeepSeek Harness 的结构化长期记忆系统
Structured long-term memory system for DeepSeek Harness.
安装
dsh plugin --profile web add github:yytbit/dsh-plugin-meta-memoryGitHub 源码安装:首次需按提示配置 allowBuilds 构建授权后重试
安装与环境配置指引、插件开发教程见 DSH 中文社区文档 ↗
安装即在你的机器上以你的权限运行第三方代码——它可读写文件、使用凭据、访问网络,DSH 的工具审批不会为插件代码加沙箱。「检测到 manifest」仅代表发现 dsh.bundle / dsh.plugin 清单,不构成兼容性或安全审查;安装前请审阅源码,不熟悉的插件先在不含密钥的环境试用。
README
目录
Structured long-term memory system for DeepSeek Harness.
What it does
Teaches the agent how to organize memories using a structured system:
- Units: Work categories named
verb-modifier-noun.unit - Records: Each memory has a brief (quick scan) and full (detailed) version
- Index: Self-maintained global storyline
- Auto-injection: Relevant briefs are injected into context automatically
Install
dsh plugin --profile your-profile add dsh-plugin-meta-memory
How it works
Memory structure
.meta-memory/
├── index.md
└── units/
├── debug-auth-module.unit/
│ ├── 0.token-refresh-fail.brief.md
│ └── 0.token-refresh-fail.full.md
└── read-deeplearning-paper.unit/
├── 0.transformer-attention.brief.md
└── 0.transformer-attention.full.md
Brief format (injected into context)
# token-refresh-fail
JWT token refresh returns 401 due to timestamp offset calculation error.
Full format (loaded on demand)
# token-refresh-fail
## Context
Auth module was failing on token refresh after 1 hour.
## Finding
expiresAt calculated using milliseconds instead of seconds.
## Evidence
src/auth/refresh.ts:42 - Math.floor(Date.now() / 1000) fixes it.
## Relevance
Any JWT/token related debugging in the future.
CLI tool
mm init # Initialize memory directory
mm list # List all units and records
mm read <unit> # Read all briefs in a unit
mm search <query> # Search briefs by keyword
mm index # Show the index
Configuration
- id: meta-memory
name: dsh-plugin-meta-memory
config:
memoryPath: '~/.dsh/meta-memory'
enableInjection: true
enableSkill: true
maxBriefBytes: 4096
License
MIT -- YYTbit
原始 README: https://github.com/YYTbit/dsh-plugin-meta-memory/blob/master/README.md ↗
同类插件
查看全部 →
mnemon
LLM监督持久内存插件 — 基于图的召回,跨会话知识,单二进制。与DeepSeek Harness、Claude Code、OpenClaw及任何代理运行时兼容。

memtrace-public
面向 AI 编码 agent 的结构化记忆:双时态图谱、MCP 原生、零 LLM 调用;支持 Cursor · Claude Code · Codex · DeepSeek Harness · Hermes · VS Code · Windsurf

dsh-flowix-memory
将本地 flowix-cli 注册为 MCP 服务,让 agent 可以搜索、读取、创建和编辑 Flowix 备忘与思维导图产物。

flowix
笔记助你,记忆助你的代理。

engramory
AI代理的便携式内存协议 — 以静态规则加载;整理学科 + 参考规范 + 可选的钩子API

dsh-memory-evolve
为 DeepSeek Harness 提供纯插件实现的跨会话长期记忆与后台自我进化能力:五轨记忆、Git 分支感知、回合内自我审查、技能自我进化与技能管理器、四轨待办、COI 调度、会话广播、会话搜索、提示词管理器和临时信息便签;零核心修改、零运行时依赖,安装即用,卸载即净。