Plugins

DeepSeek Harness out-of-tree plugin: give a text-only coding model eyes by routing images to a Qwen-VL (DashScope) route through ctx.llm and returning text.

Workspace-bound arbitrary file upload, reading, OCR, and rendering for DeepSeek Harness.

On-demand vision for text-only DeepSeek models: upload images, and the model calls a view_image tool backed by any OpenAI-compatible vision endpoint (Qwen/DashScope by default).

Local-first structured vision for text-only agents: images go to a local OpenAI-compatible VLM and come back as JSON evidence (summary, verbatim OCR, layout regions, entities/relations, colors, explicit uncertainty), with anti-hallucination fallback and an optional paste/upload bridge; zero cloud cost, images never leave the machine.

DSH Desktop toolkit with separate settings for each feature: a selection toolbar, a floating assistant, and screenshot, long screenshot and screen recording modes opened by global shortcuts from a toolbar at the top of the screen. Screenshots support shapes, pen, arrows, text and numbered markers, mosaic, pin to desktop, saving, and forwarding to a DSH conversation or the system share panel (Windows and macOS); recordings capture an area, a window or a full screen as MP4, GIF or WebM with optional system sound and microphone. Extract text, Translate and Knowledge graph on captures use the OCR tools of dsh-vision-router, which is installed separately (the settings page offers a one-click install). Requires DSH Desktop >=0.2.2-rc.1 <0.3.0-0. The features run in a separate Electron helper process and the plugin does not download Electron: with the installer build of DSH Desktop you must provide a standalone Electron through the electronExecutable setting, the DSH_DESKTOP_TOOLS_ELECTRON environment variable or $DSH_HOME/electron/; a DSH Desktop started from a bare Electron reuses it. System sound recording is not available on Linux.

Find photo candidates from the 2026 Qiandao Lake cycling race by bib number and visible gear, with local helper scripts and album snapshot comparison. Needs the event's photo album link; snapshot comparison uses Python 3.10+, bib OCR is macOS only (Swift and Apple Vision), and the optional gear ranking needs Pillow, PyTorch, Transformers and a local DINOv2 model directory.

Learn from Bilibili videos, web pages, local files and live streams: local ASR, on-screen OCR and extractive digests add up to about 1.5KB of text per source, so the model reads a summary instead of a full transcript.

Socratic tutor tools for DSH: cited retrieval, page images, chapter index, session review, embedded console. Teaching behaviour lives in dsh-socratic-persona.

Windows desktop operation layer for DSH: 17 tools for window listing, screenshot plus OCR, UI Automation, keyboard and mouse, file operations and CDP browser control. Non-intrusive by default (UI Automation patterns and window messages, never the physical pointer), irreversible steps go through a native confirmation dialog, Esc or an on-page button stops a running task, and a global lease keeps two sessions from driving the desktop at once.

Cleans a Wallpaper Engine workshop folder against a screenshot of your subscriptions: it finds the blue check badges, OCRs the tile titles, matches them to each project.json title, and moves non-subscribed wallpapers to the Recycle Bin. Ships a settings page and a floating modal, and is Windows-only.

Lightweight local visual grounding engine exposed as agent tools

Runs the local llama.cpp Qwen3.5-4B from a Settings page — Windows only, NVIDIA GPU (8 GB VRAM handles the 4B Q4) — with start, stop and restart, launch parameters, deployment paths with a presence check, provider-route and compaction-policy sync in the profile patch, a loopback request-capture proxy, and a local_worker tool that hands translation, OCR and video-frame chores to the local model by file path. Needs llama-server.exe (verified on b12069) and Qwen3.5-4B GGUF + mmproj files on disk; video frames also need ffmpeg.

macOS computer use with eight computer_* tools: accessibility-tree and Vision OCR screen reading, actions by element index or visible label that also reach background windows, and CoreGraphics mouse and keyboard events with a resident low-latency daemon for game-grade input sequences; its Swift helpers compile on first use, so there is no runtime dependency.

DSH host bundle for the ShyPie privacy-first workflow: extracts attached images and PDF/Office files locally before they enter the session log (requires poppler pdftotext/pdftoppm on PATH and a local OCR engine — Windows OCR on Windows, optionally tesseract; without them attachments are rejected, not forwarded); default filter mode keeps text local (optional local masking); blocks unmaskable media/BLOB; provides local privacy_extract, privacy_ocr, privacy_scan and privacy_rehydrate tools; denies read_image and binary/document reads outside allowPaths; and enforces a fail-closed llm/stream egress guard before DSH dispatches a model request (the default provider allowlist is the local Magpie gateway).

Computer Use for every DSH session: screen capture, mouse and keyboard, app launch, plan and skill replay, an audit log, guards for dangerous actions, a background code agent, and an OCR workflow (text and coordinates) for apps whose accessibility tree is empty. Verified on macOS only: Windows has a basic GUI backend that is unverified, Linux has no desktop backend (calls fail with unsupported_platform), and the window/AX/browser/clipboard tools register on macOS only.

Obsidian-compatible Markdown notes in the DSH sidebar with search, tags, backlinks, graph, audio notes with transcription, a paint canvas, OCR (needs tesseract), client-side encryption and notes_* agent tools; the release tarball and the npm package carry a prebuilt Linux x64 daemon, other platforms build it from source with Go.

Use your own OpenAI-compatible vision model in DeepSeek Harness: images are described before the main model sees the request, plus an image_understand tool for reading charts, screenshots, UI and OCR.

给 AI agent 一双手和一双眼睛：截屏 OCR / 按文字点 / 精细滚动 / 等文字出现 / 键鼠剪贴板 / 多步任务（Windows，DSH 插件，19 个工具）

DeepSeek Harness plugin: a desktop tool that screenshots the Windows screen (labelled coordinate grid), reads UI Automation and OCR, and drives the mouse and keyboard.

Automatically groups installed plugins in dsh-market on startup.

Pavement design methods as a judgement manual: 76 cards over 9 chapters on method selection, indicator assignment, parameter calibers, axle-load conversion and climate zoning, with the airport chapter recovered by re-OCR of the intact PDF.

Adds a global document-routing prompt so PDFs, images, and Office files are read with the right tool.

Reads the local NTQQ (QQ 9.9.x) message database on Windows and Linux and exports group and private chat history to txt, csv, json, md, html or sqlite. Ships sixteen model tools: qqnt_status and qqnt_doctor for environment and blocker checks, qqnt_extract_key, qqnt_index to build or inspect the sidecar index, qqnt_list_groups, qqnt_list_chats, qqnt_search, qqnt_stats for totals and monthly counts, qqnt_export_group, qqnt_export_chat and qqnt_export_batch (a chat can be named the way the user says it, and ambiguity refuses instead of guessing), qqnt_jobs for OCR, speech-to-text, chunking and model downloads, qqnt_media for a chat's downloaded images, qqnt_storage for disk usage and allow-listed cleanup, and qqnt_ask plus qqnt_recall for retrieval with citations. The local web UI on 127.0.0.1:46728 adds an incremental SQLite index, six views (chats, export, statistics, ask, settings, models), a task panel, a single-pane layout on narrow screens, sender redaction, media bundling, per-day splitting, export presets and AES-256 password ZIPs; OCR, speech and embedding components install into the plugin's own virtualenv on demand. Needs python3 with sqlcipher3 (the settings page can create a private virtualenv), and the database key is read from the running, logged-in QQ client.