madage/dsh-self-improved

Name: madage/dsh-self-improved
Author: madage

Plugin插件 Native原生 ⭐ 11 MIT Memory & Knowledge记忆与知识库 Prompts & Skills提示词与技能

DeepSeek 长期记忆与自进化插件：L0 捕获 -> L1 记忆提取 -> L2 场景分组 -> L3 用户画像，自动回忆注入 + 技能综合，完全本地化。

Project Overview项目介绍

Edited by Zhiwei Agent (zhiwei)由知微Agent(zhiwei)编辑

dsh-self-improved is a native plugin built exclusively for DeepSeek Harness, adding two core missing capabilities: cross-session long-term memory and agent self-evolution. It automatically distills key information like user facts, preferences, events, and instructions from conversations into a fully local SQLite store with FTS5 and sqlite-vec support. Before each new conversation turn, it retrieves and injects relevant memories into the prompt, so the AI can remember context from past interactions. It follows the four-layer memory pyramid architecture from TencentDB Agent Memory and reuses existing DSH native services, keeping all data local with no external uploads.

This plugin is designed for DeepSeek Harness users who want their agent to retain long-term user information and accumulate conversational experience over time. During regular use, it runs automatic capture, extraction, and recall injection of memories on a scheduled basis. It also consolidates similar memories, prunes outdated or irrelevant content, distills successful workflows into reusable DSH skills, and updates the user persona over the course of conversations. Users can manage their stored memory through a dedicated settings panel or the built-in /memory command line tool.

dsh-self-improved is released under the open-source MIT license, so it is free to use, modify, and distribute. It can be installed with one click from the DSH plugin marketplace, or via the npm CLI command, or directly from the GitHub source. After installation, you just need to restart DSH for the plugin to be automatically loaded, no manual configuration is required for recent versions. All development milestones are complete, all unit tests pass, and the only common caveat is a potential peer dependency double-instance issue with pnpm, which can be fixed by adjusting your pnpm store directory configuration.

dsh-self-improved 是专为 DeepSeek Harness 开发的原生长期记忆与自我进化插件，为 DSH 补上了跨会话记忆存储与自我演进的核心能力。它会自动从对话中提取要点（事实、偏好、指令等）存入本地存储，在每次新对话开始前注入相关记忆，让 AI 记住用户的相关过往信息。它遵循腾讯云 Agent Memory 的四层记忆金字塔架构，复用 DSH 原生服务，使用本地 SQLite（FTS5 + sqlite-vec）存储，所有数据都不上传第三方。

这款插件面向需要让 DSH 代理长期记住用户信息、沉淀对话经验的 DSH 用户。日常使用中，插件会自动完成对话内容捕获、关键信息提取、相关记忆检索注入，还会按调度自动整理记忆，合并相似内容，遗忘无用信息，将成功的工作流提炼为可复用的 DSH 技能，同时不断演进用户画像。用户也可以通过设置面板或 /memory 命令手动管理记忆。

插件采用 MIT 许可协议，完全开源免费，可通过 DSH 插件市场一键安装，也支持用 npm 命令、GitHub 源码等方式安装，安装后重启 DSH 即可自动加载。目前所有开发里程碑均已完成，全部单元测试都通过，需要注意 pnpm 依赖安装时可能出现的 peer 依赖双实例问题，可通过配置 pnpm 存储路径解决。

View on GitHub在 GitHub 查看

Send to your DSH发给你的 DSH 安装

请帮我安装这个 DSH 插件。安装前先完成【兼容性检查 + 安全性检查】，检查通过再动手。

插件：dsh-self-improved（madage/dsh-self-improved）
仓库：https://github.com/madage/dsh-self-improved
本站详情页：https://www.yhbd.top/plugins/madage-dsh-self-improved/
本站登记：类型 plugin · 归类 原生 DSH 插件 · 许可证 MIT · ⭐ 11 · 最近提交 2026-08-15 · 主语言 TypeScript

按下面顺序执行，每步先把结论告诉我，再进入下一步：

【1 兼容性检查】
① 我这边：DSH 版本、Node 版本、操作系统、当前 profile（web / desktop）。
② 读它的 README、package.json、插件 manifest，列出它要求的 DSH 版本 / Node 版本 / 操作系统 / 外部依赖 / 需要另外先装的运行时。
③ 逐条比对，结论只写「满足 / 不满足 / 未知」三种；不满足的给出可行替代方案。
④ 检查是否和我已装的插件冲突：命令名重复、skill / tool 重名、端口占用、重复注册的 MCP server。

【2 安全性检查】
① 仓库可信度：和上面「本站登记」是否一致；star / fork 数、创建时间、最近提交，是否归档或长期停更。
② 安装脚本：逐行看 package.json 的 preinstall / install / postinstall，以及 install.sh、setup.ps1 之类脚本。出现 curl|bash、下载后直接执行、混淆代码、访问与插件功能无关的域名，立刻停下来告诉我，不要继续装。
③ 依赖：列出新增依赖，标出无人维护、或与知名包拼写近似的可疑包（typosquatting）。
④ 权限与副作用：它会读写哪些目录、访问哪些域名、需要哪些 DSH 权限（filesystem / network / shell / clipboard 等），以及怎么卸载和回滚。
⑤ 如果它要求 sudo / 管理员权限，或权限明显超出功能所需，先停下来问我。

【3 安装】
上面两步没有「不满足」和「高危项」时才执行；用官方推荐方式安装，不要自行提权。

【4 汇报】
用表格输出：检查项 / 结论 / 依据 / 是否需要我决策。拿不准的一律写「未知」并说明要我怎么确认——不要猜，也不要替我决定。

Send this message to DSH in your current session: it verifies compatibility and security first (answering met / not met / unknown item by item) and only installs once everything checks out — it will stop and ask you if it finds a high-risk item. The box scrolls; the copy is the full prompt. CLI install commands may not be accurate across systems, so DSH is the safer route.把上面这条消息直接发给当前会话里的 DSH：它会先核对兼容性与安全性（逐条给「满足 / 不满足 / 未知」），确认没问题再安装，有高危项会停下来问你。框内可滚动，复制到的是完整提示词；安装命令不一定准确，发给 DSH 更稳。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 note1 项提示

11 stars - an early-stage project星标 11，属于早期项目

DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时（Electron / Python / Docker 等）是否齐备
Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突：命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致，是否归档或长期停更
Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需（filesystem / network / shell / clipboard）
Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限，以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照，不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装（适合开发者）

CLI Install命令行安装

dsh plugin --profile web add dsh-self-improved

把 madage/dsh-self-improved 加入你的 DSH 配置（web profile）即可启用。

READMEREADME

9.6 KB · 缓存于 2026-08-23 03:41:47 · 9.6 KB

dsh-self-improved

Long-term memory & self-evolving plugin for DeepSeek Harness (fully local).

Status: M0–M6 complete and deployed to a real environment (web profile). Design/research docs stay local only (see .gitignore).

What it is

Adds the two missing capabilities to DSH — "cross-session memory + self-evolution":

Memory: automatically distills key points from conversations (facts / preferences / events / instructions) into a local memory store; before each new turn, relevant memories are injected to the model — the AI "remembers you".
Self-evolution: memories are consolidated, decayed and corrected; successful workflows can be distilled into reusable skills; the user persona keeps evolving with conversations.

The architecture follows the four-layer memory pyramid of TencentDB Agent Memory (L0 capture → L1 extraction → L2 scene grouping → L3 persona), but reuses DSH-native services (ctx.llm / session events / agent/pre-step injection / dsh-skill / storageDomain) with a fully local SQLite store (FTS5 + sqlite-vec). No data is uploaded anywhere.

Roadmap

Milestone	Scope	Status
M0	Probe: event capture / recall injection / tool registration / settings namespace	✅ Verified (isolated headless)
M1	Memory store: SQLite + FTS5 + jieba + sqlite-vec; L0 capture to disk; memory/search tools	✅ Verified (unit + headless integration)
M2	Extraction pipeline: `ctx.llm` L1 extraction + strict JSON validation/fallback + dedup + throttled pump	✅ Unit-tested; running in production
M3	Recall injection: `agent/pre-step` injection + keyword/vector/hybrid retrieval (RRF)	✅ Unit-tested + end-to-end verified
M4	Self-evolution: L2/L3 consolidation (scenes + versioned persona), decay, correct/forget tools, skill synthesis → dsh-skill	✅ Unit-tested; synthesized skills in production
M5	UI/ops: settings panel (auto-rendered) + hot runtime toggles + `/memory` command + memory browser	✅ Complete, deployed to web profile
M6	Growth governance (caps/cleanup) + scheduling (nightly review / free maintenance / startup backfill)	✅ Complete: governance caps, nightly review (default 22:00), 15-min loop is maintenance-only, master switch stops all timers

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要，完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

渲染基于仓库 madage/dsh-self-improved 的 README.md。在 GitHub 查看原文 →