dsh-memory

DeepSeek Harness Memory Low risk

Cross-session plaintext memory: deterministic BM25 keyword recall (no embeddings), memory_save/list/search/confirm/forget tools behind a human-confirm gate, and global + project-scoped JSON stores that follow your git repo.

跨会话明文记忆：确定性 BM25 关键词召回（无向量嵌入），memory_save/list/search/confirm/forget 五个工具 + 人工确认闸门，global/project 两层 JSON 存储（project 随 git 分享）。

How to install

DeepSeek Harness dsh plugin add @max-null/dsh-memory

About

dsh-memory 一个面向 DeepSeek Harness 的**跨会话明文记忆插件**。遵循「一切皆插件」——它不修改 DSH 源码，声明 name/inject/apply，由 Loader 从 cordis.yml 加载。设计原则 1. **人是所有者**：模型只能写入 suggested 状态的记忆，绝不自我提升；只有人工确认（setStatus）才能让记忆生效。 2. **可观测先于精准**：每条记忆是明文，memory_list 随时可见、memory_forget 随时删除——不存在"静默暗礁"。 3. **明文是人机共享的审计窗口**：记忆是可读文本，模型可自检其是否过期或出错（规划中的 v2）。 4. **确定性且缓存安全**：BM25 关键词检索是存储的纯函数、无 LLM 调用；固定指引进 system-prompt section，auto 记忆进 context。用法 bash npm install @max-null/dsh-memory 在你的 cordis.yml 加一条（其余 storage / system-prompt / tools 由宿主已有；记忆的存储后端由插件自己注册）： yaml id: memory name: '@max-null/dsh-memory' 提供的服务与工具 **服务** ctx.memory：rememb…

GitHub ↗

Recommendation signals

49 Tool quality · Based on stars, downloads, maintenance, security and docs

– User interest · Adjusted by in-site views, install copies and download clicks

49 Overall

0views

0install copies

0download clicks

0outbound clicks

Links

GitHub ↗ npm ↗ Homepage ↗ Report issue ↗

Basic safety check

Findings: None
Sources: curated:awesome-dsh-plugin.com, curated:awesome-dsh-plugin/awesome-dsh-plugin
Topics: cordis, deepseek, deepseek-harness, dsh-plugin, memory

Related plugins

DeepSeek Harness New

Score97

hindsight

vectorize-io/hindsight/tree/main/hindsight-integrations/coding-agents

Hindsight, agent memory that learns: long-term project memory with auto recall and retain, knowledge pages, deep reflection, and per-repo memory banks.

☆ 20.0K ↓ 116.9K Memory ↗

MCP ServerClaude Code New

Score93

firecrawl-mcp-server

firecrawl/firecrawl-mcp-server

🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients.

☆ 7.3K ↓ 522.5K Memory ↗

Claude CodeCodex New

Score80

hivemind

activeloopai/hivemind

Hivemind turns your traces into reusable skills across agents

☆ 1.6K ↓ 4.3K Memory ↗