
Near-native image understanding for DeepSeek Harness
Install Dsh Vision
Likely fails to load: not on npm, needs a build, and ships no `prepare` script — a git install fetches sources, not built artifacts, so the entry point is never produced.
This install is expected to fail
prepare script to run it, and is not published to npm — so nothing produces the compiled output the loader then goes looking for.Three things would fix it, all of them on the author's side: publish the package to npm, add a prepare script that builds on install, or commit the build output to the repository. Until one of those happens, the command below is here so you can reproduce the failure, not so you can install it.
What would be run — expected to install and then fail to load
# 1. add the bundle to a profile — --profile my-profile points at $DSH_HOME/profiles/my-profile
dsh plugin --profile my-profile add github:oil-oil/dsh-vision
# 2. verify: the bundle should show up as a layer in the resolved config
dsh --profile my-profile --dump-config
# 3. boot the profile
dsh --profile my-profileBefore you run itinstall scripts
- This package declares
prepack— package-manager lifecycle scripts that execute on your machine at install time, outside the agent sandbox, before any tool-approval prompt exists. - A plugin runs with your permissions once loaded: it can read your files, use your credentials and reach the network. Tool approvals do not sandbox it.
- Installing from git resolves a moving branch. Pin a commit —
github:oil-oil/dsh-vision#<commit-sha>— so the code you reviewed is the code you install.
Prerequisites
- API key
- Required — see the repository README for which provider.
- Network access
- Required at runtime.
- Language
- TypeScript — a git install pulls this source, not build output.
Install check
Every field the verdict was derived from, so you can re-derive it yourself
| Field checked | Result | Weight | What the spec says about it |
|---|---|---|---|
| lifecycleScripts | fail | warn | runs at install time, outside the agent sandbox: prepack |
| dsh.bundle | pass | fatal | package.json declares a `dsh.bundle` layer |
| dsh.bundle.patch file | pass | fatal | patch file present at `cordis.patch.yml` |
| npm registry | missing | warn | `@oil-oil/dsh-vision` is not published to npm |
| prepare script | missing | fatal | needs a build but ships no `prepare` script — a git install fetches sources, not built artifacts |
- Verdict reason
- git-only + needs build + no prepare script
- package.json read
- package.json
- Rules applied from
- docs/user/develop/basic/publish.md
- Checked at
- 2026-08-15
- Install-time scripts
prepack
From the README
Excerpt as published by the author, plain text, unedited
English | 中文 `dsh-vision` is a plugin for DeepSeek Harness. Vision-capable models keep receiving images natively. When the selected main model is text-only, the plugin asks a separate vision model to observe the original images, then lets the original DeepSeek model produce the final answer. | Main model | Image path | Final answer | | --- | --- | --- | | Supports images | Original images are sent directly, without preprocessing or OCR | Current model | | `deepseek-official` or another text-only model | A configured vision model observes the original images; its output is injected as untrusted attachment context | DeepSeek | | Cloud vision unavailable | Falls back to macOS Vision or Tesseract | DeepSeek | The plugin does not replace the main model selected in H
Topics
GitHub topics on this repository
More in Image Understanding Qwen Mm Plugins
Ranked by similarity inside the cluster, not alphabetically
Dsh Vision Complete
Yts1919
给 DeepSeek 补上「眼睛和耳朵」的多模态视觉插件:看图 / OCR / 物体检测 / 视频理解 / 语音转写 / 截图直读,一键安装(DSH 插件)。
Dsh Vision Proxy
Flyvhidbwo
DeepSeek Harness 插件:DeepSeek 大脑 + 自动识图。GUI 附加图片自动经 OpenAI 兼容 VLM 转译成文字后交给 DeepSeek 作答;支持百炼/智谱/OpenRouter 等任意 OpenAI 兼容端点(默认 qwen3.7-flash),无 key 自动探测本地 Ollama(图片不出本机);安装时有一问式确认
Dsh Qwen Mm
RRRosmontis
Qwen-MM-Plugins integration bundle for DeepSeek Harness (dsh) — multimodal MCP tools (vision, OCR, ASR, search, video, Blender, FreeCAD) + image attachment bridge. 让 DeepSeek Harness 原生支持多模态。
Shadow Vision
WardLu
Open-source MCP vision server that gives text-only LLMs and AI agents image understanding, OCR, visual analysis, UI inspection, and multimodal capabilities.
Dsh Qwen Mm Plugins
ShuiHan268
DeepSeek Harness 的 Qwen-MM-Plugins 集成插件:12 个多模态 MCP 工具(视觉/OCR/定位/ASR/音视频)、Web 设置页(粘贴 Qwen API Key 即用)、内置技能与一键安装器
Dsh Plugin Deepeye
Favio8
DeepEye vision plugin for DeepSeek Harness (DSH): image description, OCR, VQA, UI layout, and clipboard analysis.