Plugins

Galatea: shape an AI assistant of your own. One assistant with its own persona and long-term memory (git-versioned plain files); long chats hand off to a fresh conversation without forgetting, and it tidies its memory in the background while you are away. It runs on whatever model DSH is set to and keeps its memory when you switch; every module can be turned off, and other plugins can extend it. Phone chat (QQ, WeChat) and Claude Code sessions come from the separate Galatea channels and agent-sdk plugins. Preview; needs DSH 0.2.2-rc.1 or a later 0.2.x.

Gives each new DeepSeek Harness conversation started from the default workspace its own workspace directory and sandbox root (manually added project workspaces stay shared unless mode is set to all), with a composer toggle for the next conversation.

Automatically split long DSH conversations into volumes (every 10 user messages by default), with todo handoff snapshots, source navigation, read-only archives and host compatibility checks.

Conversation rewind for DSH: rewind in place to any earlier user message with one click or the /rewind and /undo commands, without a new branch or window, and optionally restore workspace files edited by write tools. A Settings page lists recent rewinds per workspace and can undo the file changes from the latest one. Derived from SiriLee/dsh-rewind and maintained separately; do not install both (they share the /rewind and /undo commands and the rewind-snapshots directory).

Per-workspace Fork library: a bookmark on each finished assistant turn stores that turn boundary as a full compressed snapshot that survives deletion of the source session; a "Fork 库" conversation tab lists a workspace's forks as a tree (a fork saved from a launched session nests under its parent) with name, turn count, last question and answer, note and source link, and "从这里启动新会话" seeds a new session from any fork, repeatable. Each fork then names itself from its own context: the snapshot is replayed to the model with a naming request appended, and that exchange is never written into the snapshot or a session.

Context control for DeepSeek Harness: inspect token usage across request and conversation; switch skills, MCP servers and plugins between loaded, brief and hidden per session; toggle prompt sections and tools; compact on demand or at a configurable 5–95% threshold.

A deliberately context-free chat panel in the right sidebar: it sends only a one-line system prompt plus what you type there, with no tools, no session history and no fork of the current conversation, and keeps the panel history in memory. Each reply reports its own token usage, plus the cache hit rate on routes that return cache stats.

Turns the DSH Web UI into a macOS-style desktop with the chat parked behind it: Sonoma wallpaper, menu bar, magnifying Dock, F4 Launchpad, ⌘Space Spotlight, Control Centre, clock/calendar widgets, F3 Mission Control and a desktop orb that floats the native conversation back up. Finder, Terminal (a /bin/bash session) and Settings ship with the plugin; the Browser, Code, Office and Notes windows are unlocked by separate companion plugins (dsh-ego-browser, dsh-code-workbench, dsh-univer-office, dsh-obsidian-notes) and show a placeholder until installed, and the Browser window also launches a headless Chromium (found under /usr/bin, or set DSH_MACOS_CHROME). The agent can drive the desktop through /api/macos-desktop/* routes and window.__macos.

Codex conversation panel for DeepSeek Harness: lists local Codex threads, renders their history, and sends messages back to the original thread through the Codex CLI. Needs the Codex CLI and Python installed locally.

Suggested next-step prompts under every finished agent turn in the DeepSeek Harness chat: the model proposes a few followup prompts at the end of each turn, they render as clickable cards under that turn, and clicking one sends it verbatim as your next message. Each card stays attached to the turn that produced it across reloads and restarts, the generation language can be chosen from 15 languages or follow the conversation, and a settings page in the dsh settings covers the language, how many suggestions per turn, the header and the in-app update check. Requires DeepSeek Harness 0.2.x (web profile) and Node.js 20+.

Adds a DSH settings panel for QQ proactive-message dice rolls, showing per-conversation progress, hit statistics and quiet/guarantee countdowns, with editable parameters that apply immediately. Reads from an auto-detected local bridge, a custom HTTP status endpoint, or a built-in demo engine.

Real session deletion for DSH: trash by default, plan before delete, audit log, residue receipt

A local literature workbench for Windows x64 with PDF reading and annotations, automatic initial reading, separate per-paper Harness conversations, and a library assistant that searches metadata and shared summaries.

Pulls your chat.deepseek.com conversations into a local knowledge base — one Markdown file per conversation, a JSONL transcript and a catalog — and searches them from a control page inside DSH. Daily increments plus a full-history backfill from the official export, an integrity check, and an optional model step that runs through the host llm service and writes only after its JSON passes validation. Reads the web userToken from the DSH credential store and calls chat.deepseek.com only; no install scripts, no runtime dependencies.

Keeps prompt caches warm for OpenCode Go DeepSeek conversations by predicting each request's cache hit before it is sent, rotating the x-opencode-session routing id when that request would be cold and recent requests are slow, and pooling idle warm ids for reuse by new conversations.

Continue DeepSeek Harness conversations on an Android watch and install the watch app with a desktop wizard.

Slack-style shared rooms with independent native DSH sessions, parallel member mentions, bounded agent handoffs, text attachments and conversation export.

Every conversation gets an ending: mark dev-done, release channel and version in the conversation header, see how many items still need attention in the sidebar, and read all unfinished conversations, tasks and schedules in the right-panel ledger.

Workspace-wide conversation tree for DeepSeek Harness: every conversation and branch on one canvas with its source turn, own token usage and notes; compare two branches, bring a summary back, export a Markdown outline, archive (undoable through DSH's native archive), and delete a whole subtree (macOS and Windows only; deletion runs immediately and is permanent).

Conversation map for the DeepSeek Harness web GUI: one page per workspace, conversation cards placed freely, links between cards, named and colored boxes that group conversations, high/medium/low priority marks, a picker for which conversations (archived ones included) are on the map, and a file-based API so another conversation can rearrange the map.

Star conversation exchanges and edit any sent user message in the native composer while keeping previous conversation versions. Supports DSH 0.2.2-rc.1.

Injects a per-conversation x-opencode-session header for OpenCode Go routes that were registered under a custom provider name, through the host official llm-pi-ai/request-headers waterfall. Zero dependencies, no fetch patching, and it never overwrites a header someone else already set.

A local small-model habit hint plugin for DeepSeek Harness that reads recent turns, extracts the user habits, key points and hints, and injects them into the conversation system prompt. It needs a Python virtualenv with torch/transformers/peft plus a user-supplied Qwen3-0.6B base model and LoRA adapter, none of which are bundled; its default scope only runs in sessions that companion-memory has enabled, and with the shipped placeholder paths it silently injects nothing.

Hides the signed-in deepseek-account model group from the model picker and rewrites an already-saved account selection to deepseek-official in prepareCall and stream, so a stale conversation cannot spend on the account; no settings page and no runtime dependencies. Requires an official API key configured under Settings → Models first — without one there is no model left to select after the account group is hidden.