Full vision-capability bundle for DeepSeek Harness: a vision_understand tool (OpenAI-compatible vision APIs, free Zhipu GLM-4V-Flash by default) plus paste/drag-and-drop/button entry points for image recognition.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-vision-tools
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:moon09300731/dsh-vision-tools
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
This plugin publishes its README in Chinese only.
DeepSeek Harness(DSH)视觉能力全家桶 —— 让 DeepSeek 纯文本模型"看得见"。
- vision_understand 工具:调用 OpenAI 兼容视觉大模型 API 理解本地图片(描述画面、识别文字、回答问题),注册为全局工具,所有会话可用。
- 三入口识图:
Cmd/Ctrl+V粘贴截图、拖图到按钮、点按钮选文件 → 图片自动落盘到$DSH_HOME/pasted-images/→ 输入框填入请识别这张图片:<路径>→ 发送后模型自动调用识图工具。
默认使用智谱 GLM-4.6V-Flash(免费),支持 4 家 provider 切换。被限流时**自动降级到 GLM-4V(glm-4v-flash)**重试,免费模型高峰期也不容易失败。
安装
# 方式一:npm 安装(推荐)
dsh plugin --profile web add dsh-vision-tools
# 方式二:GitHub 安装
dsh plugin --profile web add "github:moon09300731/dsh-vision-tools#main"
重启 dsh web 后生效。
配置(vision_understand 工具需要)
创建 ~/.dsh/vision.env(全局生效,推荐):
VISION_PROVIDER=zhipu # zhipu | dashscope | siliconflow | openai
VISION_API_KEY=你的APIKey
可选覆盖:
VISION_BASE_URL=https://open.bigmodel.cn/api/paas/v4/chat/completions
VISION_MODEL=glm-4.6v-flash
限流自动降级(可选):
VISION_FALLBACK_MODEL=glm-4v-flash
- 主模型
VISION_MODEL被限流(HTTP 429 / 负载过高 / 频率限制等)时,自动降级到VISION_FALLBACK_MODEL重试一次 - 缺省降级模型 = provider 预设模型(
zhipu→glm-4v-flash,即 v4 版本);主模型与降级模型相同时不会重复请求 - 仅限流类错误触发降级;密钥无效、参数错误等业务错误不降级,直接报错
| provider | 默认模型 | 说明 |
|---|---|---|
zhipu |
glm-4.6v-flash |
智谱,免费(128K 上下文,支持思考模式) |
dashscope |
qwen-vl-plus |
阿里百炼 |
siliconflow |
Qwen/Qwen2.5-VL-7B-Instruct |
硅基流动 |
openai |
gpt-4o-mini |
OpenAI |
工作区回退:在项目目录放 .dsh-vision.env(同格式),仅该项目生效。配置每次调用实时读取,改完无需重启。
使用
- 粘贴:直接
Cmd/Ctrl+V粘贴剪贴板截图(捕获阶段拦截,优先于 GUI 自身附件处理) - 拖拽:拖图片到输入框左侧的「📷 识图」按钮
- 选择:点「📷 识图」按钮选文件
发送后 agent 会自动调用 vision_understand 识别图片。
安全边界
- 图片会经外部视觉 API 出网(base64 传输),敏感数据请改用本地 OCR(tesseract/paddleocr),或自建内网视觉服务用
VISION_BASE_URL指向 vision.env含 API Key,不要提交到 git(本仓库 .gitignore 已忽略)- 粘贴路由仅监听 DSH 本机端口,图片保存于
$DSH_HOME/pasted-images/
技术说明
⚠️ 依赖约定:
@deepseek-ai/dsh-tools是 DSH 宿主运行时自带(bundle 机制提供),本插件不声明为 dependencies。若声明,dsh plugin add触发 npm install 会在 profile 里装出第二份 dsh-tools,与宿主全局那份形成模块双实例,导致工具执行层scheduler.prepare崩溃(Cannot read properties of undefined (reading 'prepare'))。安装后建议确认 profile 的node_modules/@deepseek-ai/dsh-tools是符号链接或单实例。
- 宿主半区:
src/index.mjs(Cordis 插件,inject: ['tools', 'webServer'])vision_understand工具经defineTool注册(@deepseek-ai/dsh-tools)POST /api/vision-paste路由落盘粘贴图片
- 浏览器半区:
lib/client.bundle.js(手写__ModuleLoader__bundle,零构建依赖)conversation.input.leftslot 注册「📷 识图」按钮- 捕获阶段
paste监听拦截剪贴板图片
License
MIT
Links
More in this category
liustack/modlens★ 4100
Vision bridge for text-only models: paste an image, get structured JSON evidence (OCR, layout, semantics).
ysr666/dsh-vision-router★ 1128
Free vision for text-only agents: built-in keyless vision chain plus pixel tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots); paste an image to use it.
Anionex/dsh-vision-toolkit★ 887
Vision for text-only models: paste an image and the model switches to a Vision Toolkit variant for image Q&A, multi-image comparison, long-screenshot OCR, screenshot-to-UI reproduction, element grounding, and pixel diff. No API key by default — images are processed by the author-hosted free service, 100 per machine per day; configurable to your own provider.
dickpy/dsh-imagegen★ 99
AI image generation for the DSH Web GUI: text-to-image and image-to-image through a configurable OpenAI-compatible endpoint (gpt-image-2 / gpt-image-1 / dall-e-3), with an api_url/api_key settings card and a sidebar split-pane generation studio.
fandc520/dsh-comfyui★ 94
Drive a local or remote ComfyUI server from DeepSeek Harness: comfyui_run / comfyui_object_info / comfyui_workflow tools generate and edit images and videos, with a workflow library (graph extraction: per component / main flow / all), a load area with resolution auto-match, a live queue, SDXL and Wan 2.1 templates, a companion skill, and a same-origin media proxy.
sunxin-ai/dsh-design-qa★ 44
Design-fidelity QA for text-only models: a `deepseek_vision` tool borrows an eye from any OpenAI-compatible vision route, so the model can judge whether an implementation matches its mock — shipped with the benchmark behind that judgement (four fixtures, 23 injected defects, raw transcripts) and the questioning discipline it depends on.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.