Plugins

Workspace checkpoints per tool call: records every file mutation inside the workspace (excluding .git and node_modules by default), including changes made by shell commands rather than by write/edit, and stores the previous content of each changed path in a content-addressed store kept outside the workspace. Restores have two modes — undo one record, or return to a point in time — with a line-level diff preview, a dry run by default, and a rescue record written before every restore so the restore itself can be undone.

Model-callable compact_agents tool that force-compacts context in every live session (main session, ordinary sub-agents and AgentTeams members alike), queues busy targets and compacts them when the turn ends, reports the shadowed node count and estimated tokens per target, and ships a browser form for the trigger ratio, retained ratio, controlled-phase output budget, auto-continue budget and an end-of-turn pre-compaction ratio that compacts near the trigger line once a turn ends, so a reply no longer opens with a wait for the summarizer.

Long-form novel writing plugin: each book is a git repository of frontmatter files, with 39 DSH domain tools and a web writing UI. Deep import drafts candidates, AI adoption into canonical assets requires DSH approval, and deterministic continuity checks ship with a report-only benchmark. Independent open source (MIT), not an official DeepSeek product; requires Node >= 24.11.0 and currently targets DSH 0.1.2-rc.1.

Layered memory for DSH whose agent-facing surface is 10 tools: high-privilege writes (memory_add, memory_delete, memory_import), rumination controls (memory_ruminate, _cancel, _status), and a memory graph (memory_search_graph, memory_expand_graph_node), built over an L0-L3 distillation pipeline that recalls and injects relevant memories before each model step.

dsh (DeepSeek Harness) 跨 agent 信箱插件：让 dsh 会话内的 agent 通过 agent-mailbox 与本机其他 agent（Claude Code / OpenCode / Hermes…）互发消息、收任务卡、卡动即唤醒。零 npm 依赖，Node 22 即装即用。

Ollama 监控与测评平台：node 半侧提供 DSH 工具，web 半侧提供设置页内的交互面板

Opens the dsh web GUI in a dedicated Edge/Chrome app-mode window on its own browser profile and gracefully stops the server when that window is closed.

Per-session question index for the DSH Web UI: a time-proportional dot rail on the right edge and an expandable chain panel showing every user question with its time, including branch/subagent sessions; clicking a node jumps to that message and highlights it.

Share the DSH workspace over the local network: phones, tablets and other computers browse, download and upload files via browser with no DSH login. Token gate, path boundary, read-only mode and rate limit. Zero-dependency (node:http).

Browse local image, video, and SVGA artifacts from conversation nodes in a Finder-style column view.

Raises Node's HTTP headers/body timeouts for the whole DSH process (default 30 min, proxy-aware) so Ollama and LM Studio models that stay silent while thinking or generating a large tool call are not cut off at 5 minutes; replaces the node_modules undici patch.

Clean Architecture reference library for DSH: 8 pinned reference repos (Go, Java, C#, Python) indexed into a local FTS5 database, with assertion rules, a review ledger, and a read-only observation panel.

科研收敛驾驶舱 (developer preview): hard-gated research convergence loop for DSH agents - 7 conv_* tools + host session projection + turn-tail card (conversation.chat.turnTail) + cockpit strip (conversation.input.dock) + settings section. 零 npm 依赖 (node builtins

Conversation history panel: per-exchange delete, in-place restore and a built-in node rail, with no leftover tool-call cards after deletion.

SOP-driven business Agent Teams for DeepSeek Harness: configurable captain/member agents, human- or agent-defined SOP workflows (nodes, dependencies, quality gates, manual gates), event-driven task scheduling, and a web dashboard (SOP canvas, run board, a

Workflow One - Orchestrateur visuel de flux de travail IA pour DeepSeek Harness (dsh), avec DAG multi-agents, exécution en direct, récupération et intégration Feishu

Système de mémoire léger pour l'interface web DSH : quatre fichiers Markdown en majuscules (USER.md / PROJECT.md / WORKLOG.md / CONVENTION.md) + deux actions (append pour le flux / distill pour la sédimentation). Zéro dépendance externe, uniquement des modules intégrés à Node ; fichiers visualisables, comparables avec diff et gérables avec git.

RPC bidirectionnel entre hôte DSH et client via la passerelle native Connection et Typert, avec appels pairs ciblés, annulation et gestion du cycle de vie des connexions.

Plugin de vérification de mise à jour DSH (bundle statique) : à la première ouverture, vérifie automatiquement les nouvelles versions de DSH (prend la version npm la plus haute réelle), affiche une invite en bas à gauche avec un bouton [Mettre à jour vers] qui exécute une mise à niveau déjà vérifiée ; l'UI utilise les tokens de thème sémantiques DSH ; la racine d'installation / version de DSH, node/npm sont détectés à l'exécution, sans chemins codés en dur.

Plugin DeepSeek Harness (dsh) pour @tabilo/ai-code-review : enregistre l'outil ai_code_review (review | context) qui réutilise le moteur Node.js pour collecter des diffs, exécuter tests/couverture, faire l'analyse d'impact, et alimenter le modèle en compétences pack.

L'outil run_kernel_code pour DeepSeek Harness : noyaux Python et JavaScript persistants (un sous-processus python3/node de longue durée par session, partageant un pilote hôte central) — un plugin autonome ne nécessitant aucune modification du harnais en amont.

Répondeur d'approbations DeepSeek Harness (dsh) adossé au sidecar agent-governance — chaque approbation/demande est tranchée par la porte de refus par défaut, les règles L0–L4 et la chaîne d'audit organisationnelle.

Plugin de terminal local pour l'interface web de DeepSeek Harness : un panneau xterm.js multi-onglets sur de vrais shells node-pty, diffusé via une route WebSocket dans le processus hôte. Connectable à chaud — monté via la liste de bundles du profil web, sans modifier le code source de dsh.

Détection de la pile technologique du projet et recommandations de capacités pour DeepSeek Harness.