GooDAnDReaDY/dsh-vision-bridge

Universal vision bridge for DeepSeek Harness: attachments with native models, 40+ tools, PDF/OCR/diagrams.

Project Overview项目介绍

@goodandready/dsh-vision-bridge is a native Cordis plugin built exclusively for DeepSeek Harness, shipped as the scoped npm package @goodandready/dsh-vision-bridge and mounted inside the Cordis runtime. It targets the strict server-side modality check in DSH 0.1.2-alpha.2+ that rejects image attachments when the active conversation model lacks image in inputModalities, decorating ctx.llm.resolveModelInfo and ctx.llm.listModels so the session controller accepts images. The plugin then intercepts image blocks at the agent/pre-step and llm/stream hooks, routes them to a configured vision backend (DSH catalog provider, OpenAI-compatible endpoint, Ollama, or Webhook), and rewrites them as text markers before the text-only chat model sees the prompt.

The intended workflow starts in the DSH Web UI: a user attaches an image, the bridge first checks whether the chat model supports vision natively and lets it pass through untouched, otherwise it generates a description plus OCR, substitutes the block with a [The user attached an image. Description: ...] marker, and forwards the enriched prompt. Beyond background rewriting, roughly forty explicit vision tools are exposed for OCR (including a local Tesseract engine), visual question answering, bounding-box grounding, document/table/formula extraction, QR and barcode reading, UI-flow reconstruction, and an opt-in vision_consensus mode gated by the consensusEnabled setting, making it useful for DSH users who want multimodal power without changing chat models.

Requirements are Node.js 20 or newer plus system binaries poppler (so pdftotext actually ships the PDF text layer) and ffmpeg for frame extraction; the CI installs poppler without sudo and runs once per commit, and a vision_consensus opt-in plus warning surfaces appear in every tool schema. License is MIT, the test suite reports 329 passing tests across 104 suites, and v0.6.0 ships a one-click in-app updater endpoint /api/dsh-vision-bridge/update plus a fail-closed settings validator enforcing loopback origin, CSRF, and bearer-token checks, so first-run users should verify the scoped package identity and ensure poppler plus ffmpeg are present before enabling image attachments.

@goodandready/dsh-vision-bridge 是一个面向 DeepSeek Harness(DSH)的原生 Cordis 插件,作为发布在 npm 上的作用域包安装,专门解决纯文本聊天模型无法接收图像附件的问题。它通过装饰 ctx.llm.resolveModelInfo 与 ctx.llm.listModels、并在 agent/pre-step 与 llm/stream 钩子中改写图片块,让 DSH 会话控制器接受任意模型上的图像输入,将其转交给配置的视觉模型(DSH 目录、OpenAI 兼容端点、Ollama 或 Webhook)生成描述后再回写为文本上下文。

典型工作流:用户在 DSH Web UI 拖入图片,桥接层先判断聊天模型是否原生支持视觉,是则原图直传,否则调用视觉通道得到 OCR 与描述,注入到提示词中。对外暴露约 40 个专用工具——OCR(含本地 Tesseract)、VQA、边界框定位、文档/表格/公式抽取、二维码识别、UI 流程重建、多模型共识(vision_consensus 由 consensusEnabled 开启)——供 agent 按需调用。

依赖与运行要求:Node.js 20+,系统需安装 poppler(用于 pdftotext PDF 文本层)与 ffmpeg(帧抽取),OCR 工具链为本地 Tesseract;MIT 许可,全量 329 项测试覆盖 104 个套件,v0.6.0 起内置单点自动更新端点与失败关闭的设置校验,部署时需注意作用域包名一致性。

Pre-install check安装前体检Compatibility · Security兼容性 · 安全性 1 warning1 项注意
  • Only 2 stars - very few users, little community feedback星标只有 2,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项

Compatibility兼容性

  • DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
  • External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
  • Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册

Security安全性

  • Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
  • Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
  • curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
  • Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
  • Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
  • Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式

Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。

Or use CLI install (for developers)或使用命令行安装(适合开发者)

CLI Install命令行安装

dsh plugin --profile web add @goodandready/dsh-vision-bridge

把 GooDAnDReaDY/dsh-vision-bridge 加入你的 DSH 配置(web profile)即可启用。

READMEREADME

📦 @goodandready/dsh-vision-bridge

Universal Vision Bridge for DeepSeek Harness: Seamless Multimodal Attachments with Text-Only Chat Models

npm version license DSH Plugin Node version

GoodAndReady Showcase

🇬🇧 English • 🇷🇺 Русский • 🇨🇳 中文说明

⭐ If you like this plugin, please star it on GitHub — it shows me that the plugin is useful to you and motivates me to keep developing it.

🐛 If you find a bug or would like to request a feature, open a GitHub issue in any language — I will review your proposal and implement useful suggestions in a future plugin version.

⚡ Overview & The Problem

When interacting with text-only chat models (any provider/model that accepts text only) in DeepSeek Harness, users cannot natively attach and send images:

  1. In DSH 0.1.2-alpha.2+, the core session controller performs a strict server-side modality check (ctx.llm.resolveModelInfo). If the active conversation model lacks image in inputModalities, the prompt is immediately rejected with a session/attachment-invalid error ("Model does not support image input").
  2. Standard text-only adapters throw errors when encountering raw multimodal image blocks in their message payload.

Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →

← 上一个 Prev dsh-desktop-operator 下一个 Next dsh-capture →