플러그인
DeepSeek-Harness 플러그인을 찾아보고, 필터링하고, 설치하세요.
플러그인 367개 찾음
dsh-vision-bridge
ximengxiaolan/dsh-vision-bridge
컴포저에 첨부된 이미지를 텍스트 전용 DeepSeek 모델에 전달하기 전에 OpenAI 호환 비전 모델이 텍스트로 변환
dsh-manga-workflow
yidawuyi/dsh-manga-workflow
Manga translation pipeline for DeepSeek Harness: sidebar workbench with box types (body/SFX/title) and per-box GPU erase, local Japanese/Korean/English OCR with vision-engine fallback, DSH-model translation, a reviewed "confirm embed" flow with preflight checks and page previews, staged DSH takeover modes (SFX recognition, whole-book localization), page jump, version checks, and an automatic box-detection benchmark.
dsh-looptree-dashboard
roshenn/dsh-looptree-dashboard
Milestone dashboard that keeps a project's intent on disk next to the code: one swimlane per project showing project vision → milestone → capability → execution records, a candidate-capability pool with reasoning, per-node structured prompts previewed before they are sent, DSH sessions launched from the node that needs them with plan/execute/reflect loops, and the live status of the dev services behind the project. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.
novel-craft-dsh
sunlaoshi-123/novel-craft-dsh
Long-form novel writing skill: premise, structure, timeline tools, chapter self-check and revision accounting, featuring 17 reference documents and 5 local scripts.
paper-review-revision
songrui111/paper-review-revision
Adds a four-round review of an engineering or applied-mathematics paper to a DeepSeek Harness session: academic review, line-by-line verification of the maths and theorem proofs, English editing that returns both the original and the revised sentence, and a final pass over the revision against the baseline. Reads multi-file LaTeX projects, DOCX and PDF; every finding keeps a stable ID with its severity, evidence and handling status, and the rounds can run alone or as the full sequence.
dsh-comfyui-link
jx70/dsh-comfyui-link
Nine agent tools connect DeepSeek Harness to a local ComfyUI server for text-to-image, image-to-image, custom API workflows, LoRA presets with prompt documents, algorithm upscaling, and review-and-redraw with a vision model; the bundle also adds a Draw agent preset with generation-model and LoRA pickers beside the composer.
dsh-eye
chddydwd/dsh-eye
Local screen ring for DeepSeek Harness: keeps the last ~45 seconds of screen frames in memory so the model can rewind what just happened, recalls any frame at 1:1 pixels, reads window text through the Windows accessibility API instead of OCR, locates controls by name and reports coordinates only (never clicks), and masks private regions in every capture. Windows only; everything stays on the machine. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.
dsh-qibi
huajikano/dsh-qibi
QiBi writing desk: a native sidebar panel and 12 qibi_* tools for a local SQLite manuscript library, with optimistic-lock saves, pre-write revision backups, BM25 search, whole-book export, and model-assisted prose, character extraction and continuity notes.
dsh-model-capability-detector
bunnysakura/dsh-model-capability-detector
Fills model context, output, vision and reasoning settings for third-party API providers in DSH.
dsh-vision-direct
eurekame2000/dsh-vision-direct
Self-contained vision for text-only DSH models: registers its own vision_read_image tool (verbatim OCR, reading-order layout, semantics, uncertainty), turns pasted images into local file paths through a capture-phase paste channel with rolling storage, and optionally exposes a loopback OpenAI/Anthropic-compatible endpoint. Talks to your own upstream directly, with no external vision CLI and no runtime dependency.
dsh-video-vision
listendh/dsh-video-vision
Give DeepSeek Harness eyes for video: samples frames one per detected shot, delivers them as a labelled contact sheet, and reads subtitle text plus container metadata.
dsh-video-insight
azerz/dsh-video-insight
Video analysis panel in the DeepSeek Harness sidebar. Paste a YouTube or Bilibili link to fetch its subtitles and generate a summary, chapters, a mind map and citation-backed Q&A over the transcript, with an embedded player whose timestamps jump to that moment and keep playing. Videos without subtitles fall back to local Whisper / SenseVoice transcription (audio stays on the machine); silent videos fall back to frame OCR with macOS Vision, or to vision-model frame descriptions. Per-video analysis caching, a 100-item searchable history, and two agent tools: video_digest and video. A Bilibili SESSDATA cookie, if you enter one, is stored in ~/.dsh/video-credentials.json (mode 0600), transcripts are sent to your current model, and its HTTP routes have no DSH access check.
dsh-plugins (dsh-plan-tab-autoclose)
feng-yifan/dsh-plugins
After you approve a plan review, automatically close every plan tab of that session in the right sidebar, including tabs left by earlier revision rounds.
dsh-formula-editor
yaozhang0224/dsh-formula-editor
Image-to-LaTeX formula editor that works out of the box: it reads a dropped or pasted screenshot with the vision model your DSH already has configured, so there is no endpoint, API key or account to set up. A MathLive field edits the formula in place with a 272-symbol palette, and the result copies into Word as a native equation or exports to .docx. Any model that accepts image input works, and you can pin one specific model in the plugin settings instead of following the current default.
dsh-docmate
miaojinqwq/dsh-docmate
Immersive reading desk: md / docx / pdf render in the centre pane, and the right column is one ordinary Harness session per document that has already been fed the whole text, the research memory and the existing highlights. Select a sentence to ask about it; scanned PDFs and images have no text layer, so a marquee drag captures that region and sends it to a vision model under a turn-scoped runtime brake (max 2 steps, per-step output cap, every tool call denied before dispatch, 45 s cancel). Highlights are anchored as character offsets and painted with the CSS Custom Highlight API, so they survive reloads. Also KaTeX math, outline, in-document search and a draggable three-column layout. Zero runtime dependencies: pdf.js and KaTeX are vendored.
dsh-super-image
byton-patrica/dsh-super-image
Generates images in parallel with two configurable OpenAI Images-compatible or ComfyUI backends, compares the candidates, and optionally fuses them. Visual comparison needs a vision-capable model in DSH; without one the candidates are listed unscored.
dsh-matlab
tedcriss/dsh-matlab
Connector for MathWorks' official MATLAB Agentic Toolkit skills and the MATLAB MCP Server. Requires a local MATLAB R2021a or later. Skill groups can be enabled individually on the settings page. The toolkit is provided by MathWorks' official repository https://github.com/matlab/matlab-agentic-toolkit; the trigger rate for MATLAB-related operations is over 90%, and it calls the native MCP server, so your AI works in a fully standardized way. Its HTTP routes have no DSH access check, so a web page open in your browser could call them. By default (autoProvision) it downloads the MathWorks MCP Server binary and the newest Agentic Toolkit release on first use; MathWorks telemetry is disabled by default.
dsh-supervisor
harzva/dsh-supervisor
Per-session supervision with role assignments and handoff workflows; off until enabled with /supervise. Requires @harzva/dsh-role-registry 0.1.0-alpha.1 (not on npm) installed first from its GitHub tag archive.
dsh-video-replica
limomss-dunang/dsh-video-replica
Six-step video-replica workbench: fetches a Douyin/Xiaohongshu/WeChat Channels/Kuaishou/Bilibili share link or a local file, breaks the source into shots with a vision model and optional ASR, writes three A/B/C storyboards, generates a first frame and a clip per shot, then stitches the final video with optional TTS narration, burned-in ASS subtitles and BGM. Needs Node 22.18+/24.11+, ffmpeg and ffprobe on PATH, and configured vision/chat/image/video model endpoints; share links go through a third-party parser service (or yt-dlp), or upload a local file.
dsh-latex-formula-editor
jiahui-huang1/dsh-latex-formula-editor
Free LaTeX formula editor in the DSH sidebar with offline MathJax/PDF.js rendering: ten symbol tiles with hover palettes, ten template groups, live preview, colour/font/size/environment controls, favourites with titles and search, and nine export formats (SVG/PNG/JPG, LaTeX/AsciiMath/Typst/HTML/MathML/OMML) with escaping and share links. Image and PDF OCR need a vision-capable model in DSH or your own OpenAI-compatible vision endpoint; everything else runs offline.
dsh-moss-draw (plugin)
boltblue121/dsh-moss-draw
7개의 art_* 에이전트 도구와 오른쪽 사이드바 드로잉 작업대(시리즈, 페이지, 리비전, 영역, 카메라 블록아웃)를 추가합니다. 사용자가 호스팅하는 ComfyUI 인스턴스에 ComfyUI 그래프를 제출하여 MOSS/K2 텍스트 이미지 생성, 인페인트, 캐릭터 참조, 영역 편집, Qwen-Image 편집을 수행하며 요청마다 호출자 LoRA를 최대 4개 쌓을 수 있습니다. 로컬 checkout과 어댑터 단계가 필요합니다. 마켓플레이스 설치 후 install/adapt.mjs를 실행하거나(Windows는 install/install.ps1) __MOSS_*__ 자리표시자를 절대 경로 workspaceRoot와 ComfyUI origin으로 바꾸세요. 그전까지 Host는 MOSS_CONFIG_*로 프리셋 로드를 거부합니다.
deepseek-memory
rocofish/deepseek-memory
dsh용 장기 메모리를 심층 방어로 강화합니다. 프로젝트 `.dsh-memory/`와 사용자 `$DSH_HOME/memory/` 두 계층에 항목 수와 태그가 제한된 직접 편집 가능한 JSON을 저장합니다. 고유 임시 파일에서 rename하는 원자적 쓰기, 제한된 `history/`의 이전 버전, 계층별 프로세스 잠금과 오래된 잠금 회수를 지원합니다. 디스크 버전이 달라 업데이트 손실이 우려되면 덮어쓰지 않고 거부합니다. 기억 텍스트는 명시적으로 신뢰할 수 없는 데이터 경계 안에 두고 저장 시 위조 경계를 제거하며 로드 시 범위를 검증하며 계층 밖으로 나가는 심볼릭 링크는 MEMORY_STORE_ESCAPE로 차단합니다. 내용 휴리스틱 필터나 확인 단계가 없어 헤드리스 자동 통합을 유지합니다. 설치: `dsh plugin --profile web add github:RocoFish/deepseek-memory`。
dsh-dont-stop
tcy07/dsh-dont-stop
에이전트가 오랫동안 계속 작업하도록 합니다. 기한을 설정하고 일찍 끝내면 Dont-Stop이 자동으로 반복 작업과 다음 단계 수행을 권합니다.
dsh-moa-native
ddl095/dsh-moa-native
생성 시점 네이티브 Mixture-of-Agents. 매 턴 참조 모델들이 같은 문맥으로 병렬 초안을 만들고 통합 모델 하나가 전체 도구로 실행합니다. 모델형 설정 레시피, F 미러/R 축약 보기, 이중 형태 세션 색인의 하위 Agent 제어, 레시피별 vision(이미지 참조가 텍스트 전용 통합 모델에 판독 중계)을 제공합니다. 조향·순찰 명단 분리 주기와 사이 구간 고정 재사용, 퇴화 출력 제외, 공급자 익명 실패 표시. 예행 기록을 생성 위치에 고정해 요청마다 붙이지 않아 캐시 접두사를 재사용하고 각 단계는 추가분만 과금하며 모든 주기에서 기본 활성입니다.