DeepSeek Harness plugin

dsh-agent-memory-findshan

File-based self-evolving memory for DeepSeek Harness v2: compression & extraction over official compaction, daily episodic layer, dream consolidation to semantic files (user/agent/memory/dream +

Jump to install

Source facts

Repository
findshan/dsh-agent-memory
Latest update
Aug 16, 2026
Category
Workflow & Automation
GitHub stars
0
Format
plugin
Catalog evidence
Upstream dsh.bundle evidence
Evidence path
package.json#dsh.bundle
Checked against
0.1.0-rc.8
Upstream check date
2026-08-20

This evidence comes from the upstream catalog. This site has not installed, run, or security-reviewed the plugin.

Install

Start with a prompt that asks an agent to review the GitHub repository and source. Switch to the command if you want to install it yourself.

Copy this prompt into DSH, Codex, or another agent and ask it to review the GitHub repository and source first.

Do not install or run any commands yet. Read this plugin's GitHub repository, README, and relevant source code. Then answer the questions below clearly and directly so I can decide whether it fits my needs:

1. What is this plugin, and what problem does it solve?
2. Who is it for, and what are its typical use cases?
3. How is it used after installation? Include one minimal example.
4. What known limitations or privacy, security, compatibility, or maintenance risks does it have?
5. Give a clear recommendation: recommend, conditionally recommend, or do not recommend, with reasons.

Distinguish statements documented by the repository, inferences from source code, and unknowns. If evidence is insufficient, say so explicitly. Do not guess or simply repeat the README.

GitHub: https://github.com/findshan/dsh-agent-memory
Plugin: dsh-agent-memory-findshan
Author: findshan

Check the source files

Read the README and other files from this plugin directory before installing.

File explorer3 files
README.mdSource · read only

dsh-evolving-memory

文件化自进化记忆 —— DeepSeek Harness 的跨会话压缩层(v2)

记忆的本质是信息的压缩与提取。本插件叠在官方 compaction(会话内压缩)之上,用廉价模型把已压缩的会话摘要进一步提取、整合为人类可读的 Markdown 记忆文件——用户可看、可改,智能体可检索、可进化。

设计要点

记忆 = 压缩与提取消费官方 compaction/summary 事件做提取(T1),不重复读原始日志;dream 做跨会话整合(T3)
纯 Markdown 文件user / agent / memory / dream + 每项目 project.md + daily/ 时间层;无 schema、无迁移、无置信度、无状态机
人机共治user.md 用户可读可编辑;agent 写入只是建议,用户有最终决定权
目录式披露注入常驻注入只有记忆目录(确定性生成,零模型调用,有界);内容按需 memory_search(找)+ memory_read(展开)披露——与 DSH skill 机制同构
三层压缩会话 compact(官方)→ 提取进 daily/(情景记忆)→ dream 固化进主题文件(语义记忆),对应"睡眠时情景→语义固化"
廉价模型驱动提取与整合走便宜模型(KB 级输入一次调用);无 key 时记忆读写/检索照常,仅模型步骤跳过

Install / 安装

dsh plugin --profile web add dsh-evolving-memory
dsh --profile web

记忆文件落在 $DSH_HOME/memory/

├── daily/2026-08-16.md   # 今日要点 + 会话纪要 + 待办(时间层,dream 后归档)
├── user.md               # 用户画像(身份/偏好/目标/禁忌/想法)—— 用户可编辑
├── agent.md              # 智能体自我认知与工作方式
├── memory.md             # 长期记忆(兜底落点)
├── dream.md              # 梦境整合日志(审计)
└── projects/<p>/project.md

工具(6 个)

memory_search(找)· memory_read(展开)· memory_catalog(目录)· memory_save(写)· memory_correct(纠错即学)· memory_dream(整合)

分类是软约定(写进 prompt,不是代码):用户的事→user.md,项目的事→project.md,时间→daily/,拿不准→memory.md。检索兜底一切——放错文件也能找到。

配置

Key默认含义
memoryDir$DSH_HOME/memory记忆根目录
dreamIntervalHours / dreamMinSessions24 / 5梦境整合门控
model / apiKey / baseURLdeepseek-chat / env / api.deepseek.com廉价模型(提取+整合)
catalogBudgetTokens / catalogTopN1000 / 5目录注入预算与每文件行数
searchTopK5默认检索条数
autoExtracttrue消费 compact summary 自动提取
dailyRetentionDays30daily 保留天数,dream 后归档

Development / 开发

npm run typecheck
npm run build
node test/smoke.mjs          # 无模型:布局/保存/读取/检索/目录/纠错/捕获/时间层/持久化
DEEPSEEK_API_KEY=sk-... node test/integration.mjs  # 真实模型:提取/纠错/梦境整合

为什么做这个

DSH 生态的记忆插件要么是 JSON 记录 + 置信度机器(复杂),要么没有跨会话整合。v2 回归本质:记忆就是几份活的文档 + 一个会读写的智能体——系统只提供文件、检索器、整合触发器,判断全部交给模型。明文本是 AGI/ASI 时代的通用接口。PRD:[PRD.md](PRD.md) · 旧方案:[PRD-v1.md](PRD-v1.md)

License

MIT