AI Agent Hub
Back to plugins
🖥️

dsh-vision-toggle

Client Updated 2026.09.04

Run the following command in DeepSeek Harness:

dsh plugin install lijian-ui/dsh-vision-toggle

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

Install the dsh-vision-toggle plugin (source: https://github.com/lijian-ui/dsh-vision-toggle), then restart the dsh desktop to enable or disable image input per model from the settings page.

About this plugin

The dsh desktop settings page offers no way to declare input modalities for custom provider models. If you connect a vision-capable third-party model through an OpenAI-compatible service such as LM Studio, vLLM, or Ollama, the only workaround today is to hand-edit settings.yaml and add the input field for each model, a fragile process that is easy to get wrong.

This plugin adds a Model Vision section to the settings page. Every configured model is listed, grouped by provider, each with a Support image toggle. Flipping it on writes [text, image] and flipping it off reverts to [text], persisted through the official llm-pi-ai settings namespace so the change takes effect instantly with no restart. If the host write fails, the switch rolls back automatically and an error is shown.

Built for users who connect vision-capable models and would rather click a switch than dig through configuration documentation, one click and image input is ready to go.

Use Cases

  • Connecting a vision-capable model via LM Studio or vLLM and attaching images in chat
  • Managing multiple models where only some support vision, toggling individually
  • Enabling or disabling image input from the settings page instead of editing YAML

Best For

  • Advanced dsh desktop users connecting third-party vision models
  • Researchers who frequently switch between text-only and image input
  • Everyday users who prefer GUI toggles over manual config edits