Plugins

Design-first PowerPoint builder: six pptx_* tools turn a deck.json design draft into an editable .pptx with AI backgrounds, cutout illustrations and decoration pieces, run layout and contrast checks, render slide previews, and review pages with a configurable vision model; chat cards and a settings page in the Web UI. Requires DSH Desktop 0.2.1-rc.2+ (pptx_render uses its bundled LibreOffice Kit) and an Agnes or ModelScope API key for image generation.

Document review workbench: select text in a Markdown file, leave comments saved as MRSF sidecar files next to it, and let the agent read them with doc_review_* tools and propose edits you accept or reject.

Vision relay for text-only models: a configurable GLM-compatible vision endpoint analyzes the images you attach in the background and the original model keeps answering in the same session, leaving your message, image, model and history untouched. Requires applying patches/0001-session-image-admission.patch to a stock DSH install (images are rejected before storage otherwise); the repo ships an apply/restore script.

All-media (video/audio/image) sibling of liustack/modlens: turns local media into structured evidence (timeline / transcript / layout regions) injected into the conversation so text-only models can cite concrete frames, seconds and speakers — Evidence, not imagination. One failover chain per media, recorded hop-by-hop in meta.attempts (a fallback is never silent): host-registered multimodal models first, then a configured OpenAI-compatible endpoint (comma-separated multi-key rotation), then local ffmpeg frame-sampling / a local ASR CLI, degrading to metadata-only as a last resort. Absorbs modlens battle-tested rules — cooldown keyed only on auth/quota failures and softened to chain-tail (never dropped), redact-before-clip error text, magic-byte sniff that never trusts file extensions. Optional vision-bridge registers a media-lens wrapper into the host model selector so a pure-text model gains image/audio/video sight. Registering / authenticating multimodal models and true inline audio-video upload lean on two companion plugins, declared as optional peers (absence auto-degrades, never errors): dsh-thinking-levels fronts the model config / registration surface, dsh-llm-openai-completions takes over the OpenAI-compatible transport that serializes file blocks into video_url / input_audio. Privacy: media goes to remote providers by default; strictLocalOnly keeps content on-machine. Prerequisites not shipped with the package: ffmpeg/ffprobe on PATH (frame sampling / reencode / duration probe; degrades to metadata when missing) and a user-configured local ASR CLI for the audio fallback. Hosts 0.1.7 / 0.2.0 / 0.2.1 (6 rc, exact engines.dsh enum). npm: dsh-media-lens.

Registers an `image2_generate` tool that calls your own OpenAI-compatible image relay to generate or edit pictures (text-to-image via POST /images/generations, image-to-image via POST /images/edits with a local PNG/JPEG/WebP reference, accepting b64_json or image URLs), saves them in the session working directory, and renders them inline in the conversation through its `@wisdoverse/dsh-inline-media-viewer` dependency.

Lets DSH drive the Chrome tab you are already browsing, no vision needed: the page reaches the model as structured text with a numbered element list, and clicks, typing, key presses, scrolling, tab handling and navigation wait for your confirmation in the extension side panel while password and payment-card fields stay masked. Fifteen browser_* tools cover snapshot, text, tab list, open, follow, click, type, press, scroll, wait, navigate, back, close, forward and reload. A repack of the bridge plugin from omdsh-dev/dsh-browser, published to npm so it installs with `dsh plugin add`: the peer ranges the host checks are widened so the install is not refused as incompatible, the bundled extension also probes the DSH Desktop host port 19387 so no bridge address has to be typed, and a script copies the extension to a fixed directory so a profile switch or reinstall cannot break it. The prebuilt release tarball carries the plugin and the built MV3 extension. Upstream: https://github.com/omdsh-dev/dsh-browser.

Web search and fetch provider for DeepSeek Harness and Pi: anonymous-first routing across Exa, Tavily, Parallel, Firecrawl and TinyFish, with opt-in billing, quota preflight and bounded cancellation.

Embeds the DeepSeek web chat (chat.deepseek.com) in a DSH Desktop main panel through an Electron webview guest, with a back/forward/reload toolbar and a local snapshot of login-related localStorage restored on the next open.

Audits Chinese copy for AI-sounding phrasing: a 0-100 score over six dimensions, sentence-level flags with rewrite hints, and per-genre weights adjustable in conversation.

On-screen test-progress HUD for DSH — a click-through panel drawn over the app under test, showing each step's expectation vs actual and whether the waiting human may take back the mouse and keyboard. Ships a universal prebuilt binary (arm64 + x86_64), so no Xcode Command Line Tools are needed; the panel itself requires macOS 14+.

Requires the Codex CLI installed and `codex login` completed. Consult the OpenAI Codex CLI from DeepSeek Harness with three tools: codex_consult runs Codex in a read-only sandbox, inlines evidence files, verifies the served model and reasoning effort from the Codex rollout file, retries disconnected streams, and records quota before and after the call; codex_quota reports the current 5-hour and weekly quota; codex_calls lists recent calls.

Export a whole working DSH home — profile, installed plugins with their node_modules, skills, hooks and task boards — into one .dshpkg tar and import it on another machine without network or pnpm; also streams over HTTP or a one-shot LAN link.

One settings page for every MCP server on your machine. It reads the real config file of 17 other AI tools (Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, OpenCode, Trae, Trae CN, Qoder, Gemini CLI, Qwen, Zed, Roo Code, Copilot CLI, Kiro, CodeBuddy, Continue) plus DeepSeek Harness itself, groups servers by the file that declares them, probes each one over JSON-RPC to see whether it starts and which tools it exposes, and gives every server a switch that mounts or unmounts it in DeepSeek Harness without editing the other tools' own files.

Read-only CSV/XLSX comparison by business keys, with offline review decisions, comments and source-verified feedback handoffs.

Runs the official Google Antigravity CLI (agy) as a one-shot coding subagent over print-mode stream-json, exposed as the subagent_antigravity delegation tool. Includes Gemini 4 Argon on Antigravity subscriptions. Requires the Antigravity CLI (agy 1.4+) installed and signed in.

Convert PDF, Word, PowerPoint, Excel, HTML, CSV, JSON and ZIP files to Markdown for the model, using a bundled JavaScript engine or Microsoft MarkItDown.

Connects an existing local CodeGraph CLI to each session's indexed project, shares one MCP process per project, and updates session-scoped tools when the working directory or index changes.

Semantic code search for DeepSeek Harness: registers two native tools, semble_search (find code by describing its behaviour, returning ranked file:line hits) and semble_find_related (neighbours of a known location), which run the Semble CLI via uvx, plus one bundled routing skill. Requires uv on PATH; the first query on a repository builds its index.

123 Cloud Drive (123pan) client: signs in with a phone number or email plus password, or a pasted token, browses and drag-drop uploads and downloads through a sidebar panel, backs up a workspace incrementally while keeping its directory tree, and exposes 15 tools for an agent to list, upload, download, rename, move and delete files. Network, MD5, 16MB multipart upload and token encryption run in a bundled statically linked C++ core, so it needs no Python or other runtime. Windows 10/11 x64 only.

Keeps bulk output out of the conversation: ctk_execute, ctk_execute_file and ctk_batch_execute run code in a child process and return only the conclusion, ctk_index and ctk_search file long material into a local FTS5 knowledge base built on node:sqlite, and a routing gate refuses curl and wget before they execute. Ships ctk_stats and ctk_doctor plus a settings page with a live context-occupancy panel. No third-party runtime dependencies.

Native Windows desktop control for DSH: window enumeration, UIA control tree, in-process GDI window capture, offline OCR text finding and clicking, background UIA invocation without moving the cursor, clipboard-based CJK typing and key chords, plus a course-playback helper. Ten tools, no driver, no extra runtime.

Connects the Drafta notes app to DeepSeek Harness over MCP, so the agent can search, read and write notes in your local Drafta library.

Workspace memory for DeepSeek Harness ported from ZCode: one Markdown memory directory per workspace (a MEMORY.md index plus one file per fact), injected into each main session and refreshed by a background extraction request after completed turns; a sidebar page browses the stored files read-only.

Tide Memory: a long-term memory that forgets on purpose. It keeps what matters and lets the rest fade. Whether an entry enters the context is decided by an evidence score C = s/(s+2): below the threshold (0.35 by default) it never enters. Events decay by half-life (30 days, or 7 days when they carry emotion); facts never decay; superseded entries are kept and marked stale instead of deleted. Recalled lines do enter the context and therefore do consume tokens — small, but real. No embeddings, no network calls, no cloud.