dsh-memfuse：DSH 本地优先的长期记忆插件
dsh-memfuse 是维护者 a1141171521 为 DeepSeek Harness 开发的长期记忆插件。真源为 ~/.dsh/memory/ 下纯文本 Markdown，读写作注入与本地 CJK bigram+BM25 检索不调用模型；自动抽取可关，失败会写入注入与 status.md。本文基于官方 README 介绍存储布局、遗忘链、Agent 工具及 profile/cordis 两种安装方式与「设置→记忆」入口。
Read Morememory-skill：为 AI Agent 提供可编辑的本地长期记忆
memory-skill 是 SkillHub 社区 Skill（@user_4005c3e1/memory-skill，v1.0.0），为 AI Agent 提供基于文件系统的长期记忆：用户偏好、项目背景、错误教训、决策历史、会话摘要与调研归档。记忆默认存放在工作区 .workbuddy/memory/，支持关键词检索、Qwen3-Rerank 精排与 YAML 元数据，本地可读可改。本文介绍其定位、核心能力，以及通过 skillhub CLI 安装到 Cursor 等 Agent skills 目录的步骤与使用注意。
Read More@yyfather/dsh-memory：为 DSH 提供跨会话长期记忆与协作能力
@yyfather/dsh-memory 是 DeepSeek Harness 的记忆类插件，fork 自 csyangwen/dsh-memory-evolve 并适配 DSH 0.2.0-rc.2。提供五轨长期记忆（用户档案、全局事实、项目关键记忆、项目/每日日志）、四轨待办、技能与提示词注入、git 记忆同步、书签与任意轮分支、多会话协作房间，以及 Kimi/Codex/Grok/Hermes 等外部 CLI 异步派单。安装命令为 dsh plugin --profile web add @yyfather/dsh-memory，bundle patch 自动注册。适合需要跨会话保留项目上下文、多设备同步记忆或在 DSH 内编排内外 AI 协作的开发者；使用前请审阅源码与 MIT 许可。
Read MoreAgent心理学报告：用心理学框架做 Agent 对话协作复盘
Agent心理学报告是 SkillHub 社区 Skill（@user_1793b15c/agent-psychology，v1.0.2），在对话中基于当前上下文，用认知、情绪、社会、人格、发展、语言与工程心理学等框架诊断人机协作卡点，输出红黄绿可视化复盘报告，并将正向协作约定写入长期记忆。无需 API Key。安装需 SkillHub CLI，用 skillhub install 并指定 Agent 的 skills 目录；在会话中说「生成复盘报告」「对话诊断」等触发词即可。适合需要系统化复盘 Agent 协作的开发者。
Read Moreagent+obsidian=长期记忆：用 Obsidian 给 Agent 做跨会话记忆
介绍 SkillHub 社区 Skill「agent+obsidian=长期记忆」（@user_7da98a98/obsidian-memory，v1.0.2）。从 SOUL.md 的 OBSIDIAN_MEMORY_PATH 定位 Obsidian 记忆库，任务初加载 profile、rules 与近期日志，在「任务完成」或 /记一下 等触发词下写回 Markdown。包内 v2.0 说明含命中打点、/审计记忆、重复踩坑升级 rules.md 等 Anti-Rot 治理。含 skillhub 安装命令、记忆目录结构与 SOUL 配置要点，适合需要可编辑长期记忆的写作与开发场景。
Read Moredsh-memory：DSH 两层渐进式长期记忆
dsh-memory 是 jipika 为 DeepSeek Harness 维护的记忆类插件：全局 topics 分片与按工作区切换的项目层均以纯 markdown 存储，系统提示词仅注入实时片级地图，条目通过 memory_search 与 read 按需加载。插件不调用模型，支持设置面板编辑、三档注入方式与写入守卫。安装命令为 dsh plugin --profile web add @jipika/dsh-memory。
Read More长期记忆 / 知识图谱 会思考的进化 AI：为 Agent 搭统一记忆底座
SkillHub 社区 Skill「长期记忆 / 知识图谱 会思考的进化 AI」（@user_1a470ba8/zwjh-skill，v2.6.0）面向 AI Agent，提供跨会话长期记忆、实体关系图谱、自动沉淀，以及语义与时间线检索、记忆健康度审计和本地备份（可接百度网盘）。标注 requires_api_key=false，强调纯本地与零密钥。文章说明其解决上下文断裂的痛点，列出官方核实功能，并给出基于 skillhub CLI 在 Cursor、Claude Code、Codex 等目录下的安装命令与 SkillHub 安装文档链接。
Read More永久记忆自动归档系统：对话落盘、语义检索与跨会话继承
介绍 SkillHub 社区技能 @user_6bedba2d/ybdsjy（v1.0.1）：面向 AI Agent 的永久记忆归档，宣称每次对话自动记录、语义搜索、跨会话继承，重启不丢且不覆盖旧记录。文章说明技能定位、核实后的核心能力、skillhub CLI 与 Cursor/Claude Code 等 skills 目录安装方式，以及公开页未给出细粒度用法时的启用注意。requires_api_key=false。
Read MoreDeepSeek Harness 记忆插件：dsh-memory-dreaming
# dsh-memory-dreaming 插件：将对话转化为长期记忆 **前言**：在 DeepSeek Harness (DSH) 的开发与使用中，会话历史膨胀导致检索困难是常见痛点。`dsh-memory-dreaming` 插件专为解决这一问题设计，它不依赖 Channel Core，直接在本地维护结构化记忆，将验证过的对话转化为日常日记、梦境整合、每周复盘和可检索的长期事实。 **核心功能**： * **增量式处理**：每次处理最多 7 条新日记，回溯 30 天，仅处理归档副本，不重扫原始对话。 * **日常与复盘**：自动整理 `memory/YYYY-MM-DD.md` 并定期合
Read MoreDeepSeek Harness 角色设定与长期记忆管理插件 (dsh-plugin-persona-memory)
DeepSeek Harness（DSH）的 `dsh-plugin-persona-memory` 插件旨在解决传统记忆管理方案仅对特定预设生效的问题。它采用插件化架构，将角色设定与长期记忆注册在宿主组合层，实现全局统一管理。插件提供可视化的右侧栏管理面板、全局与项目两级记忆以及长期日志（JOURNAL.l）管理。它支持通过 `memory` 和 `persona` 工具进行读写操作，并针对不同会话（主会话与子代理）设置了差异化的注入逻辑。安装后，数据存储于特定目录，但在卸载时不会自动清理，需注意版本兼容性及更新操作的限制。
Read MoreAI Agent记忆系统：本地优先的跨会话长期记忆 Skill
AI Agent记忆系统（@user_7db0c006/agent-local-memory-keeper，v5.2.1）是 SkillHub 上的 ai-agent 类 Skill，面向跨会话失忆与结论过时问题。对话中的踩坑、纠偏与验证结论可自动沉淀；库内去重、纠错、冲突仲裁与定期自检；支持自然语言语义召回（本地中文模型、离线、数据不出本机）、版本回滚与时间点回溯。安装需 skillhub CLI 并用 --dir 指向 Cursor 等 Agent 的 skills 目录。
Read MoreDeepSeek Harness Curated Long-Term Memory Plugin
deepseek-memory is a long-term memory plugin for DeepSeek Harness (DSH). Through plain-text storage and dynamic injection, it addresses the challenge of persisting key facts in conversation history. It provides two-level storage (user-level and project-level) for structured facts, and includes tools for adding, deleting, querying, updating, and searching memories. The plugin dynamically injects memories into each request and integrates them after turns executed by a restricted sub-agent. Users need to adjust the integration mode and entry limits through configuration, and restart the configuration after installation to apply the changes.
Read Moredsh-memory-lite: Giving Your dsh Agent Cross-Session Long-Term Memory
The DSH plugin `alanzhao0128/dsh-memory-lite` provides dsh agents with zero-dependency, Markdown-based long-term memory across sessions. It persists preferences and decisions using a resident memory directory, five explicit memory tools (read, write, update, and forget), and a background implicit extraction mechanism. The plugin supports peer isolation and cross-peer sharing, with permission-based security controls and runtime status indicators. After installation, restart to use it; it is suitable for development scenarios requiring context retention and multi-session collaboration.
Read MoreDeepSeek Harness Long-Term Memory Plugin: dsh-unified-memory
`dsh-unified-memory` is a plugin that provides low-cost long-term memory for DeepSeek Harness. It uses SQLite full-text indexing to retrieve memories stored in local Markdown files only when needed, avoiding the need to load the entire memory store into the context and thereby saving tokens. The plugin supports three working modes: `dsh` (standalone storage), `overlay` (reads Claude Code memories), and `shared` (shares memories with Claude Code). After installation, users can switch modes using the `mem mode` command. All data is fully local, with no network requests, making it suitable for scenarios with limited context windows or long-running agent tasks. Users should be aware of environment dependencies and privacy concerns related to plaintext storage.
Read Morewhale-persona: Persona Engine and Long-Term Memory Plugin
whale-persona is an open-source memory plugin for DeepSeek Harness, designed to address the issue of AI frequently forgetting personas and contracts. It is blank by default and supports managing self-identification, stance, image, tone, working agreements, and chain-of-thought language through configuration files. Its core mechanism is a mandatory human-confirmation mode, ensuring that long-term memory only takes effect after user approval. Installation requires Node.js 20 and DSH 0.1.6 or later, and it supports managing presets via command-line tools and previewing injected content. The plugin is built as a pure ESM module, is secure and controllable, and does not connect to the network.
Read Moredsh-hindsight-memory: A Pipeline-Level Long-Term Memory Plugin for DeepSeek Harness
The `dsh-hindsight-memory` plugin provides pipeline-level long-term memory for DeepSeek Harness (DSH). By leveraging DSH's hook mechanism, it automatically injects and persists memories before and after model calls, enabling new sessions to automatically “learn” background context. Core features include automatic recall and persistence, fault isolation, and layered configuration. After installation, users can configure parameters such as the API endpoint and memory store name via the settings panel. Operationally, the plugin automatically retrieves relevant memories at the start of each conversation, asynchronously persists content when a conversation ends, and supports memory transfer between sub-agents. It addresses the limitation of traditional agent session memory being confined to the current context.
Read MoreDeepSeek Harness Long-Term Memory Plugin: dsh-memgas
`dsh-memgas` is a long-term memory plugin for DeepSeek Harness (DSH), designed to address context loss in multi-session interactions. The plugin is open-sourced under the MIT license and offers core capabilities including multi-channel retrieval fusion (lexical, vector, granularity, and graph) as well as a closed-loop memory evolution mechanism. It uses local SQLite for storage, supports zero-configuration installation, and provides tools and user commands such as `memory_save` and `memory_search`, enabling the intelligent injection of historical conventions into new sessions. It is suitable for scenarios that require maintaining project context across sessions and for privacy-sensitive use cases.
Read Moredsh-plugins: DeepSeek Harness Plugin Collection
The DeepSeek Harness (DSH) plugin package integrates five high-frequency plugins to address pain points in agent development. Core features include long-term memory and session archiving, Git-based session management, session branch tree management, an automatic Goal resumption mechanism, and fine-grained editing of model reasoning capabilities. The repository offers one-click installation or installation of individual plugins, supporting local development and offline operation. When installing, avoid ID conflicts and duplicate mounting; uninstallation does not delete saved data.
Read Moredsh-memory: A Long-Term Memory Plugin for DeepSeek Harness
dsh-memory is a long-term memory plugin for DeepSeek Harness (dsh), designed to address the context window limitations of agents and the fragmentation issue in conversational memory. It uses persistent JSON storage, automatically guides the model to call `memory_recall` before responding to retrieve relevant background information, and calls `memory_save` after responding to store new facts. The plugin provides five core tools (create, read, update, delete, and search), supports a 1-5 star importance rating, and includes a visual web-based management interface, intelligent search, and import/export capabilities. It can inject key memories into the context, effectively improving the memory coherence and interaction quality of agents.
Read Moredsh-remember: A Cross-Session Long-Term Memory Plugin for DeepSeek Harness
`dsh-remember` is a local memory plugin for DeepSeek Harness, designed to address agent context fragmentation. It stores experiences as local Markdown files without relying on external vector databases. The plugin offers features such as automatic memory extraction, consolidation, recall, and manual management. Through a Git workspace, users can view, search, and manage memory assets, with support for one-click translation and packaged import/export. It is suitable for developers who need to persist experiences locally and prioritize data privacy.
Read Moredsh-engram: DeepSeek Harness Cross-Session Long-Term Memory Plugin
`dsh-engram` is a cross-session memory plugin for DeepSeek Harness (DSH), designed to resolve the context gap between Agent sessions. It adopts a "Memory Palace" architecture and builds its storage system based on information architecture principles, incorporating fixed routes, skeleton reuse, and a 0-1 scoring mechanism. Core features include two-tier repositories (global and project), hybrid retrieval (FTS5 and vector), spaced-repetition review using the SM-2 algorithm, and automatic knowledge-flywheel ingestion. The plugin has built-in security safeguards, supporting prompt injection defense, data masking, and source auditing. It also provides a web management dashboard and tools such as `engram_save/search/review` to help Agents build a stable long-term memory store.
Read Moredsh-markdown-memory: A Folder-Based Long-Term Memory Plugin for DeepSeek Harness
The dsh-markdown-memory plugin addresses the pain points of long-term memory management for agents by storing memories as plain Markdown files, without requiring an additional database or daemon. It supports YAML metadata, index files, recursive directory reading, Obsidian vault mounting, and keyword-based trigger retrieval. This approach gives users full control over their memory data and supports version control, making it ideal for developers who prioritize data autonomy and lightweight deployment.
Read Moredsh-memory-vault: A Plugin Providing Long-Term Memory Capabilities for DeepSeek Harness
`dsh-memory-vault` is a DSH plugin designed to address memory limitations and forgetting caused by relying solely on the context window. It provides agents with long-term, human-like memory through an external database. The plugin supports memory management, classification statistics, conversation logging, and historical retrieval, and uses a tiered structure (short-term, medium-term, long-term, and permanent) to automatically manage information and prevent context overflow. Users do not need to invoke it manually; simply expressing intent naturally in a conversation is enough to add or query memories. Data is stored in a local database, and memory retention parameters can be adjusted via configuration files, making it a practical tool for building long-term memory agents.
Read MoreDeepSeek Harness Long-Term Memory Plugin: dsh-gitmemo
`dsh-gitmemo` is a plugin for DeepSeek Harness aimed at addressing the long-term memory storage problem for agents. It manages memory entries using a local Git repository, providing persistence and auditability while avoiding database dependencies. The plugin exposes five core tools (read, write, search, delete, and replace) through the Cordis mechanism and is registered only with the Root Agent. It supports configuring a search limit; the environment must have the Git CLI installed and the process restarted for changes to take effect, offering agents a lightweight, local long-term memory solution.
Read MoreForeSight: A Temporal Long-Term Memory Plugin for DeepSeek Harness
ForeSight is a long-term memory plugin for DeepSeek Harness (DSH), designed to address the stale injection and time-loss issues caused by traditional static memory. It treats memory as scheduled data with temporal semantics and controls its lifecycle through two dimensions: “body” and “anchor.” The plugin supports decay, conflict resolution, and configurable policies, uses SQLite vector storage, and requires local directory configuration for data and template files. It is currently in an experimental stage and depends on Node.js ≥ 20 and a local embedding service.
Read Moredsh-persist: DeepSeek Harness Long-Term Memory Plugin
The dsh-persist plugin provides hierarchical memory storage for the DeepSeek Harness Agent via the file system, solving the “losing memory when switching conversations” problem. Core features include conversation-specific memory, cross-project sharing, bge-m3-based semantic search, and a three-state automatic recall mode. All memories are stored locally as plain-text Markdown and are human-editable. Users can read and write memories through the `memory` tool or a UI tab. Installation requires Node.js >= 22.6, and enabling semantic search requires a SiliconFlow API key. The plugin’s default data paths and routing remain unchanged, and it supports privacy controls and remote access configuration.
Read Moredsh-hypatia: A DeepSeek Harness Memory Plugin Based on the Hypatia Knowledge Graph
`dsh-hypatia` is a memory plugin for DeepSeek Harness (DSH) that leverages the Hypatia knowledge graph to provide models with long-term memory. The plugin runs the Hypatia CLI as a host process, handling complexities such as database orchestration within the plugin, while the model focuses solely on memory proposals. Its core features include recall within the same request, precise project scoping, verified writes, two-stage forgetting, and session support without Bash. Installation requires Hypatia CLI 0.1.4 and Node.js 22.5+, and can be completed via the command line. Configuration supports multiple memory presets, authorization, and recall strategies. The plugin provides tools for memory storage, retrieval, forgetting, and knowledge graph queries, but has a single-process limitation and version requirements.
Read Moredsh-engram: DeepSeek Harness Ultra-Minimal Long-Term Memory Plugin (Goal: Token Savings)
The dsh-engram plugin for DeepSeek Harness (DSH) provides a minimal long-term memory system designed to reduce context bloat in agent development by conserving tokens. Its core features include zero-LLM automatic capture (based on pattern matching and redaction of sensitive data), symbolic indexing with progressive disclosure (supporting BM25 recall, drill-down tools, and repeated-failure wake-up), and the ESR (Evidence Closed-Loop) protocol for lightweight task management, including session TODO offloading, command repair closed loops, and automatic override. By strictly limiting context length, the tool is suitable for developers who need strict token-usage control and structured task management.
Read MoreIntroducing the DSH Plugin: satan9394/dsh-knowledge-brain
`dsh-knowledge-brain` is a plugin in the DeepSeek Harness ecosystem focused on long-term memory, designed to address the high cost and high latency of traditional solutions. It can build self-connecting knowledge graphs without invoking an LLM, and offers core features such as answer citation for synthesized responses, gap analysis, data isolation, and continuous operation. The plugin is suitable for enterprise scenarios that require a stable memory layer and knowledge association; its source code security should be carefully reviewed before installation.
Read MoreDeepSeek Harness Hierarchical Long-Term Memory Plugin
`dsh-layered-memory` is a static plugin for DeepSeek Harness (DSH), designed to address the issue of agents forgetting context during multi-turn conversations. The plugin organizes memory into transient, episodic, and semantic layers, uses AI to automatically refine and deduplicate information, and supports maintaining continuity across sessions. After installation, it requires zero configuration, and models can invoke four tool functions in natural language to perform memory retrieval, writing, and committing. Plugin data is stored in local JSON files, while the management interface is provided through an HTTP API, including operations for statistics, retrieval, and create, update, and delete, ensuring data privacy and security.
Read MoreDeepSeek Harness Long-Term Memory Plugin: dsh-memory-amem
`dsh-memory-amem` is a plugin for DeepSeek Harness (DSH) designed to inject A-MEM-style long-term memory. The plugin captures user messages through a six-step pipeline—filtering, analysis, retrieval, evolution, application, and persistence—associates new information with prior memories using hybrid retrieval and LLM-based decision-making, and embeds the results into the system prompt to aid model recall. It also provides security defense mechanisms and tools such as search. On the LoCoMo benchmark, the plugin significantly improves multi-hop reasoning, enabling DeepSeek models to outperform GPT-4 in closed-book question answering.
Read Moredsh-plugin-memory: Cross-Session Long-Term Memory for DSH Agents
When developing agents using DeepSeek Harness (DSH), there is an unavoidable issue: sessions lack memory. Users' preferences, project backgrounds, and previous decisions need to be re-stated every time a new session is opened. There are two common approaches to addressing this. One is to write memory into skill files and rely on the model to load them proactively—this is a "soft guarantee"; if the model forgets to invoke it, the memory won't be loaded. The other is to perform retrieval every time a query is made—memory is always used on-demand, lacking continuous accumulation. `dsh-plugin-memory` takes a third path: forcibly injecting the memory boot block into the runtime context of the system prompt at the beginning of a session to create a "hard guarantee." The mechanism and usage of this plugin are introduced below. # What is dsh-plugin-memory
Read Moredsh-governed-memory: Governance-based Long-term Memory Plugin for DeepSeek Harness
`dsh-governed-memory` is a governance-oriented long-term memory plugin designed for DeepSeek Harness (DSH), maintained by `haoyuan-sjtu` and licensed under the MIT license. Its core purpose is to address governance issues within agent long-term memory, such as record source, permissions, scope, and revocation, by constraining the capture, promote, revoke, and retrieval processes to reduce the risk of arbitrary model writing and retrieval injection. This plugin does not provide a browser UI or HTTP endpoints; it operates solely as a trusted host plugin within the DSH process. It disables direct write access on the model side, mandating that candidate records initially enter quarantine and must undergo explicit approval and evidence verification by the trusted host before they can be...
Read Moreimemory: DSH Self-Evolving Long-Term Memory Plugin
imemory is an MIT-licensed plugin for DSH designed to address the issue of long-term information loss caused by insufficient context in multi-session scenarios. It automatically extracts and scores information worth retaining after each turn, and recalls relevant memories to inject into the context before each step via keyword pre-screening and LLM semantic re-ranking. Core features include injecting user profiles at the start of new sessions, performing conflict merging and decay eviction every 6 hours, consolidating reusable processes into `learned-*` skills, and providing 6 memory management tools and a `/imemory/` web management panel. This plugin has zero dependencies and is compatible with any profile type. Installation requires DSH version 0.1.0-rc.6 and Node >= 20. It supports manual mounting (copying files and modifying `cordis.patch.yml`) or using bun.
Read Moredsh-memory: Long-term Memory Plugin for DeepSeek Harness
dsh-memory is an MIT-licensed long-term memory plugin for DeepSeek Harness (DSH), designed to address the lack of a stable medium for user preferences, verification steps, and failure lessons in multi-session workflows. Distributed via GitHub, it utilizes structured extraction to convert reusable information from completed sessions into summaries and memory files, and automatically injects them into subsequent sessions. Core features include: debounced extraction after session completion (default 3 minutes), global cooldown integration (default 6 hours) into MEMORY.md and memory_summary.md, persistent summary injection, and the provision of four query and add tools: memory_list, memory_read, memory_search, and memory_add. Among them, memory_add is only used when...
Read Moredsh-noc-memory: A plugin to integrate long-term memory into DeepSeek Harness
dsh-noc-memory is a plugin for DeepSeek Harness (DSH), designed to address the issue of context loss after a single agent session concludes and to enable long-term memory functionality. Rather than providing a standalone memory store, this plugin connects DSH to a user-deployed Noc Memory MCP server. It primarily offers two categories of capabilities: loading long-term memory (noc_boot) and daily briefings (noc_briefing) at the start of a session; and reading, searching, creating, and updating memory entries (noc_read, noc_search, noc_create, noc_update). Using this plugin requires an environment with Node.js >= 20.0.0 and execution via `dsh plugin --p`
Read Moredsh-agent-memory: Agent-driven long-term memory for DSH
dsh-agent-memory is an MIT-licensed long-term memory plugin (v0.3) built for DeepSeek Harness (DSH), aiming to provide cross-session structured memory capabilities. It adopts a hierarchical storage architecture, dividing memory into L1 facts (preferences and environment), L2 knowledge (reusable content), L3 contexts (timeline of experiences), and a summary layer, while supporting routing for both global and workspace scopes. The plugin exposes tools such as remember, recall, memory_browse, update/forget, and memory_stats to agents, supporting multi-dimensional retrieval and time pyramid browsing. Its key features include automatic time bucket compression and integration with dsh-agent-compact, automatically consolidating che...
Read Moreinterest-memory: Add cross-session long-term memory to DeepSeek Harness
`interest-memory` is a local long-term memory service maintained by djasdh, designed to address context discontinuity between DeepSeek Harness (DSH) sessions. Built with Go, it utilizes SQLite-vec vector indexing for local storage and retrieval, supporting fully offline operation. The service integrates with DSH via the Cordis plugin `@djasdh/interest-memory-dsh-bridge`, automatically recalling relevant memories at the start of a session and pushing conversation logs at the end. Core features include evidence-based memory storage, a three-stage verification process (check/claim/conflict), automatic merging of semantically similar entries, replacement marking for old entries, and graph-based query visualization. The system provides a REST API to facilitate memory sharing among multi-agent frameworks.
Read MoreDSH ExpMem: Adds a Checkable Long-Term Memory Plugin for DeepSeek Harness
DSH ExpMem is an MIT-licensed community plugin for DeepSeek Harness, designed to address the issues of local storage and retrieval of long-term personal memory. It writes user habits, task experience, and engineering insights to local JSON files, preserves DSH Session as the raw word-for-word history layer, and does not duplicate Session events. The plugin adopts a layered memory architecture, supports prompting the Agent to retain persistent knowledge when context pressure is high, and performs retrieval sorting based on relevance, recency, and importance. Its core advantages include supporting auditable reflection (observations supported by citations, interruptible and resumable), maintaining a complete trust lifecycle from candidate to deprecated, and importing Markdown memories from Claude Code and Codex as candidate records with source evidence. The plugin runs lightweight, not...
Read Moredsh-llmwiki: Local Markdown Long-term Memory Plugin for DeepSeek Harness
dsh-llmwiki is a DSH plugin developed and maintained by chancelu, designed to address the issues of limited context windows in agent models and the lack of long-term memory. It utilizes a local Markdown vault as long-term memory storage for DeepSeek Harness, enabling continuous utilization by injecting relevant Wiki knowledge into model requests. The plugin provides `memory_search` and `memory_save` tools, supports the automatic capture of daily sessions to a specified directory, and employs three retrieval strategies: keywords, Wiki link graphs, and timestamps. It combines the Reciprocal Rank Fusion (RRF) algorithm to merge results and assembles context within the token budget. It features an LRU + TTL caching mechanism and has no runtime dependencies other than Node.js (version 22 or higher).
Read Moredeepseek-harness-memory: DSH's Claude Code-style long-term memory plugin
deepseek-harness-memory is a long-term memory plugin for the DSH community, maintained by 2303572348 and licensed under the MIT License. The plugin is designed to provide cross-session persistent memory capabilities and stores data as plain markdown files (including YAML frontmatter), fully compatible with Claude Code memory formats. Key features include: supporting user-level and workspace-level memory storage, and automatically injecting memory indices into system prompts upon session start; providing a 'memory' tool for CRUD operations on memories and importing memories from Claude Code; and supporting the mapping of multiple checkouts to unified project memory via projectAliases. The plugin can be configured via cordis.patch.y...
Read MoreAgentSoul: Local Persona and Long-term Memory Plugin for DeepSeek Harness
AgentSoul is a local plugin for DeepSeek Harness, designed to provide stable persona, identity, state, and long-term memory layers for Agents. The plugin is installed as a bundle and automatically loads globally upon startup; it requires Node.js >=22.5.0 and is licensed under the MIT License. Its core capabilities include support for four-layer persona files (SOUL, IDENTITY, USER, STATE) with write protection, as well as SQLite-based local long-term memory and distillation mechanisms that automatically fall back to JSON storage when SQLite is unavailable. The plugin features exception isolation to ensure its own failures do not interrupt the main Agent loop, and it registers six model-visible management tools, including status viewing, memory search, and manual distillation. Installation is via `dsh pl...
Read Moredsh-soul-md: Persona Cards and Long-term Memory for DeepSeek Harness
dsh-soul-md is a community plugin for DeepSeek Harness (DSH) designed to centralize the management of personas and long-term memory via a settings interface, avoiding the manual maintenance of scattered files. Its core functions include "Persona Cards" and "Long-term Memory": Persona Card content is rendered into the system prompt, supports multiple cards, switching by conversation or workspace, and allows the AI to read, write, and evolve via tools. Long-term memory provides persistent storage, is subject to size limits when injected into prompts, and prioritizes association with the currently active persona card. Installation requires Node.js >=20.18, and the new version requires DSH >=0.1.0-rc.7; users with older versions need to lock to specific versions. In terms of usage, users can create cards and specify default values on the settings page, while the AI autonomously manages memory and personas using dedicated tools. This plugin is suitable for scenarios requiring...
Read Moredsh-plugin-meta-memory: Structured Long-term Memory for DeepSeek Harness
The dsh-plugin-meta-memory is a plugin in the DSH ecosystem designed to address issues of dispersed long-term memory and context reuse. It organizes memory into a "unit + record" structure, where each record distinguishes between `brief` (summary) and `full` (complete content), supports automatically injecting relevant `brief` into the Agent context, and maintains a global `index` to organize clues. The plugin provides a CLI tool `mm` that supports operations such as initialization, list viewing, reading, searching, and index management, allowing users to name units in a `verb-modifier-noun` format, enabling hierarchical storage and on-demand loading of memory. Its main features include memory classification, quick summary scanning, full record retrieval, and automatic context injection. Configuration options include `m
Read Moredsh-task-memory: Task-Isolated Long-Term Memory for DeepSeek Harness
`dsh-task-memory` is a MIT-licensed long-term memory plugin for DeepSeek Harness, designed to address the issue of task content confusion caused by global memory. This plugin uses tasks as isolation boundaries, maintaining independent memory banks for each task to ensure that `remember`, `recall`, and `search` operations are performed only within the scope of the current task. Core features include binding specific task banks via `memory_bind_task`, providing precise read/write operations by key and lexical search based on keywords (supporting both Chinese and English), supporting memory deletion and bank management, and optionally injecting the current task's pinned or recent facts into the agent's system prompt to optimize context. The plugin requires Node.js >= 20.
Read Moredsh-memory: Integrating LingShu Long-Term Memory into DeepSeek Harness
dsh-memory is a DSH memory plugin maintained by FuRongJun-1999 that integrates the LingShu (AEIS) Spatiotemporal Memory Engine into DeepSeek Harness. It provides cross-session recall/search, a knowledge flywheel, a white-box reasoning pipeline, self-awareness, and a v3.3 role-playing engine, along with a guardrail charter and full-chain audit. Installation requires first installing aeis via pip, then adding the plugin using `dsh plugin --profile` and configuring `dbPath` and `tools` mode in `cordis.yml`. It is suitable for AGI and interpretability research-oriented developers. Before use, please review the source code and the MIT license.
Read Moredsh-memory: Providing Local Git Long-Term Memory for DSH
dsh-memory (seriousz158/dsh-memory) is a local Git long-term memory plugin for DeepSeek Harness. Memory is stored in a separate Git repository instead of the cloud, and injection is controlled via the memory.enabled setting. The settings page allows for status viewing and secure clearing, with optional idle session synchronization. Installation: dsh plugin --profile web add github:seriousz158/dsh-memory. Compatible with DSH 0.1.0-rc.6/rc.7, licensed under MIT.
Read MoreDSH Noema: Installing Viewable Long-Term Memory for DeepSeek Harness
dsh-noema is a Noema long-term memory plugin developed by ZSeven-W for DSH. It integrates local-first, non-vector memory into Agents, providing noema_* tools such as recall, retrieval, and import, along with a web-based settings page. Memories are stored in Markdown format in the ~/.agent-memory directory and can be migrated from tools like Cursor, Claude Code, and Codex, with support for service keep-alive. This article introduces its positioning, features, installation commands, and usage considerations.
Read Moredsh-memory-evolve: Equipping DeepSeek Harness with Long-Term Memory and Self-Evolution
dsh-memory-evolve is a DSH memory-type plugin maintained by csyangwen, which provides capabilities such as five-track cross-session memory, git branch awareness, four-track to-do lists, skill self-evolution, COI external AI scheduling, and session collaboration in a purely plugin-based manner, with zero core modifications and zero runtime dependencies. This article introduces its positioning, core features, installation commands, and typical usage, while also explaining community directory attributes and security considerations.
Read MoreAdd Cross-Session Long-Term Memory to DeepSeek Harness with dsh-memory-evolve
`dsh-memory-evolve` is a DeepSeek Harness memory plugin maintained by csyangwen (MIT License, JavaScript). It uses five-track memory, git branch filtering and confirmation-based writing to沉淀 project conventions, user preferences and progress into `~/.dsh/memories/`, eliminating the need to repeat everything from scratch when switching sessions. Capabilities such as skill evolution, four-track todo, COI scheduling and session broadcasting are also provided in the same repository, and most of them are disabled by default. The installation command on the directory page is `dsh plugin add github:csyangwen/dsh-memory-evolve`. The community plugin library has no official affiliation with DeepSeek / Horizon Robotics (Fang), and you should check the source code and license before installing.
Read More