AI Agent Hub
Back to plugins
🤖

dsh-vision-android

Model Inference Updated 2026.08.18

Run the following command in DeepSeek Harness:

dsh plugin install superclaude1/dsh-vision-android

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install superclaude1/dsh-vision-android in your DeepSeek Harness terminal to install this plugin; the source repository is https://github.com/superclaude1/dsh-vision-android.

About this plugin

When the main model has no multimodal input, an agent can neither understand a screenshot nor operate a mobile interface.

This plugin combines vision question answering with Android UI automation: it sends local images to an OpenAI compatible vision model, or uses mobile_look to obtain a screen description and interactive element coordinates; then it runs real taps, swipes, text input, key events, installs, and launches over adb on an emulator or physical device, while mobile_dump provides UI hierarchy coordinates for verification.

It suits developers who need agents to inspect interfaces, operate Android apps, or automate mobile testing workflows, and teams that want to add visual input and device control to local harness pipelines.

Use Cases

  • Enable an agent to read phone screenshots and perform real taps
  • Automate multi-step interaction flows in Android apps
  • Add image comprehension to a text-only model pipeline

Best For

  • Developers automating mobile app testing pipelines
  • Agent engineers integrating vision into harness workflows
  • Teams building device-level UI automation with adb