AI Agent Hub
Back to plugins
dsh-image-theme preview

dsh-image-theme

Fun & Dress-up Updated 2026.08.15

Run the following command in DeepSeek Harness:

dsh plugin install Carpon39038/dsh-image-theme

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

Run dsh plugin install Carpon39038/dsh-image-theme in a DeepSeek Harness terminal; the full source is available at https://github.com/Carpon39038/dsh-image-theme.

About this plugin

The default themes in the DeepSeek Harness web client are clean but uniform; wanting to define an entire color scheme from a personal photo or design mockup used to mean manually tweaking hex values one by one. dsh-image-theme collapses that into three quick steps: drag in an image, pick an accent, and apply.

Inside the browser the plugin downsamples the image to a 96×96 grid, samples up to 6 000 opaque pixels, converts them to CIE Lab space, and runs a deterministic k-means pass to extract five stable colors. The darkest shade becomes a translucent glass surface, the user selects one of the remaining colors as the DSH accent, and the plugin corrects text-to-accent contrast before writing over a dozen public tokens through ctx.theme.overrideTokens(), touching no host source code. The original image fills the entire window as a backdrop with adjustable dark overlay, blur, saturation, and position. Every byte stays in IndexedDB and localStorage; nothing is uploaded, and disabling the plugin preserves both the image and the settings for later.

Built for developers, designers, or any daily DSH web client user who wants a cohesive theme generated from a single wallpaper, with zero design expertise required.

Screenshots

Use Cases

  • Drag in a wallpaper and get a cohesive theme in seconds
  • Give your DSH workspace a personal visual identity
  • Customize your theme with zero data leaving the browser

Best For

  • Developers who want a personal touch in their DSH interface
  • Designers who want a single photo to define the entire UI palette
  • Privacy-first users who prefer local-only workflows