DeepSeek Harness Plugin Hub

Publish and manage complete Harness Profiles. Discover Plugins for your next setup.

Explore

PluginsPresetsDocsNews

Community

Publish a pluginContactReport an issue

Resources

Plugin Hub on GitHubDeepSeek HarnessSystem statusPrivacy notice
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

Independent and unofficial. Not affiliated with, authorized by, or endorsed by DeepSeek.

Agent Memory — DSH Plugin for DeepSeek Harness
DeepSeek Harness Plugin Hub
ProfilesPluginsCategoriesNewsDocsSign inManage Profiles
ProfilesPluginsCategoriesNewsDocsSign in
← Plugins
A

@dsh-agent/dsh-agent-memory

Agent Memory

Agent-facing long-term memory for DSH. Four narrow tools — memory_remember, memory_recall, memory_expand, memory_prune — over a human-readable NDJSON file that survives across sessions. Recall returns a compact 10-column index and never the body; fetch bodies explicitly by id. There is no automatic

The plugin will be installed here. Keep web if you are unsure.

npx -y @deepseek-ai/dsh plugin --profile web add github:TindalosKorone/dsh-agent-memory#336d2a6c32275c85fb9e91978f206d41c8753ce0
READMECompatibilityVersions

Description

Agent-facing long-term memory for DSH. Four narrow tools — memory_remember, memory_recall, memory_expand, memory_prune — over a human-readable NDJSON file that survives across sessions. Recall returns a compact 10-column index and never the body; fetch bodies explicitly by id. There is no automatic topical recall: only a one-line index and a stable rule enter context, so content requires an explicit memory_recall call. Writes are validated fail-closed, atomic, and capped by record and byte limits with score-based eviction. Zero runtime npm dependencies (only the host-provided @deepseek-ai/cordis and @deepseek-ai/dsh-tools peers).

Compatibility and provenance

Agent Memory is published as @dsh-agent/dsh-agent-memory and currently resolves to version 0.1.0. The Hub verifies its manifest and preserves the exact installation source for reproducible installs.

DSH compatibility
*
Runtime surfaces
any
Release source
github
Registry updated
10/16/2026

Versions

0.1.0stable
10/16/2026

Related plugins

Loading related plugins…

Latest
0.1.0
DSH
*
HMR
Process restart
Tree shaking
Safe tree shaking not declared
Unpacked size
Unavailable
Files
Unavailable
Surface
any
License
CC-BY-NC-SA-4.0
Source
github
GitHub
★ 0
Weekly downloads
0
Last push
10/16/2026
View source ↗
README badge

Click the badge to copy Markdown for your README.

Do you maintain this Plugin?Claim benefit · Priority security scan

Verify the GitHub repository declared in package.json to manage this listing. After you claim it, Hub will prioritize a security scan of the current version and publish the result when it passes.

Claim this Plugin →
Report an issue

Related plugins

More verified plugins in memory-context.

Hindsight Coding Agents@vectorize-io/hindsight-coding-agentsReflect-only Hindsight long-term memory for coding agents (harness-pluggable: opencode, …), with automatic background ingestion (no setup CLI).Ab Memorydsh-ab-memory--- description: "The dsh-ab-memory package: a cross-session, file-backed agent memory with five tools (remember / recall / forget / load_skill / distill) and an always-on light injection." kind: "package-reference" ---Memsearch Dsh@zilliz/memsearch-dshMemSearch plugin for DeepSeek Harness: shared markdown memory across agents, with capture, pre-step context injection, memory-recall skill, and a skill-candidate review panel.Archive Manager@michengai/dsh-archive-manager归档、恢复、收藏和永久删除会话,并诊断损坏的会话日志。 · Archive, restore, favorite, and permanently delete sessions, and diagnose damaged session logs.

README

dsh-agent-memory

给 DSH Agent 用的跨会话长期记忆:四个工具(memory_remember / memory_recall / memory_expand / memory_prune)把值得留下的结论写进一份人类可读的 NDJSON 文件,下次会话按需召回。

注意:没有「按话题自动召回」。 自动进上下文的只有一行索引(条数 + 条件规则 + 标签锚点)和一条稳定规则;记忆正文不会自动注入,要内容必须由模型显式调用 memory_recall(再按 id 调 memory_expand 取正文)。那条规则是提示,不是保证。

适用场景

  • 适合:跨会话留结论 —— 事实、教训、偏好、指路。
  • 不适合:多用户 / 多租户共享记忆;「一问就自动想起相关内容」的场景(本插件不自动召回)。

安装

按 1 → 2 → 3 走:clone 到本地 → 构建产物 → 用命令装进 profile。

  1. 克隆到本地

    git clone https://github.com/TindalosKorone/dsh-agent-memory.git
    cd dsh-agent-memory
    
  2. 构建产物

    pnpm install        # 取 TypeScript 等 devDependencies(本包运行时零依赖)
    pnpm run build      # 等价 node node_modules/typescript/bin/tsc -p .
    

    lib/ 是 tsc 产物且已随包入库;只安装、不改源码时,这一步可跳过。

  3. 安装进 profile

    dsh plugin --profile <profile> add /绝对路径/dsh-agent-memory
    

    把 <profile> 换成目标 profile(如 web)。装完重启 DSH —— 插件在 boot 时载入,不重启不会出现。

只想用、不打算改代码的话,也可以跳过 clone 直接一行装远端:dsh plugin --profile <profile> add github:TindalosKorone/dsh-agent-memory。

或者,让你的 agent 按 docs/install.md 安装 —— 那份文档写全了三种方式、profile 侧软链、路径与验证步骤。

怎么用

  1. memory_remember:写一条。kind 四选一(fact / lesson / preference / pointer),title 单行 8..120 字符,tags 1..12 个,source 非空;scope 可选(建议 project:<名>,省略 = global,非空单行 <= 64 字符)。
  2. memory_recall:按查询召回索引,每条一行 10 列 id | kind | title | tags | graph | via | rel | cov | match | score。 limit 默认 5、最大 50,且是硬显示上限;scope 可选:不传 = 不过滤(与过去完全一致),传了 = 只返回该 scope 或 global 的记录(global 永远包含),表头回显 ;scope=<值>,结构化 rows 另带 scope 字段(打印行仍是 10 列)。 match 由 rel、两个阈值与内容量闸门共同判定(复算口径见契约 §5);多样性启用时 score 不能仅由 rel/graph 复算。库规模离标定点过远时表头会提示重标,实际重跑用 pnpm run calibrate -- --write。
  3. memory_expand:拿 recall 回的 id 取正文(L2)。
  4. memory_prune:同 kind 同标题合并(留最新、累加 hits),再按 pruneScore = decay(ts)*(1+hits) 从低到高淘汰;dryRun: false 才真删。

完整列语义、打分常数、写入协议与闸门见 docs/recall-contract.md。

故障排查

装完工具不出现、插件卡在 PENDING、lib/ 没构建、注入行为不变、库在哪与怎么备份 —— 这些排查步骤已移到 docs/troubleshooting.md。

安全

  • 记忆库是本地纯文本文件,位置 $DSH_HOME/agent-memory/memory.ndjson;不含任何凭据。
  • 上限:默认 2000 条 / 4 MiB;单条默认 ≤ 总上限(maxBytes,默认 4 MiB)的 10%;写超限直接拒收(不落盘、不触发淘汰)。
  • 许可:本插件采用 CC BY-NC-SA 4.0。

上游与署名

本插件参考了 lioensky/VCPToolBox 的记忆系统设计(该项目同样采用 CC BY-NC-SA 4.0)。

链接

  • docs/glossary.md — 术语表(rel/cov/match/qTok/L1/L2 等符号的唯一定义)
  • docs/install.md — 三种安装方式(远端一行 / 本地 clone + 构建 / 交给 agent)、重启与验证
  • docs/troubleshooting.md — 故障排查(工具不出现、卡 PENDING、注入不变、库位置与备份)
  • docs/design.md — 分期实现、7 条硬不变量、两条注入机制、为什么不自动召回
  • docs/recall-contract.md — 10 列逐列含义、打分常数、写入协议与 fail-closed 闸门
  • docs/development.md — 构建、测试纪律与失败证据、脚本、环境事实
  • docs/limitations.md — 9 条诚实局限、标定出处、发布注意
  • redproof/ — 失败证据目录(每条机制一份「能让测试变红」的证据)
  • NOTICE / LICENSE