dsh-vision-bridge-skill

Plugins in this repository

• dsh-llm-vision-bridgev0.1.0

  DeepSeek adapter wrapper for the dsh web profile: describes image attachments via an OpenAI-compatible vision API (DashScope qwen-vl-max by default), then delegates the text-only request to DeepSeek. Lets the session model stay deepseek-v4-flash while pasted images are read automatically.

  package.json

  plugin/package.json

  dsh.bundle.patch

  —

  Official core packages (9)

  • @deepseek-ai/dsh-anonymous-user-id^0.1.0-rc.6shared
  • @deepseek-ai/dsh-credentials^0.1.0-rc.6shared
  • @deepseek-ai/dsh-home-paths^0.1.0-rc.6shared
  • @deepseek-ai/dsh-launch-environment^0.1.0-rc.6shared
  • @deepseek-ai/dsh-llm^0.1.0-rc.6shared
  • @deepseek-ai/dsh-llm-deepseek^0.1.0-rc.6shared
  • @deepseek-ai/dsh-settings^0.1.0-rc.6shared
  • @deepseek-ai/dsh-timeout^0.1.0-rc.6shared
  • @deepseek-ai/schemastery^0.1.0-rc.6shared
• Unnamed plugin

  package.json

  —

  dsh.bundle.patch

  templates/cordis.patch.yml

Paths are pinned to commit 0e401108606f

Overview

一键式为dsh提供看图功能，借鉴了claude-vision-skill,先让agent读取了这个skill，后面尝试使用的时候发现平台会限制ds模型直接读取图片（只能剪贴板或者路径），所以做了一个桥接功能，VLM看完图后再把内容发送给DS并输出内容