Vision & Input
dsh-ui-spec
Turns UI screenshots into implementation-grade specs: tokens, spacing scale, layout grid.
What it is
A single analyze_ui_image tool that converts a screenshot or mockup into a structured frontend spec. A deterministic geometry layer measures exact dimensions, palette, suggested design tokens, layout grid and spacing scale; an optional vision model adds semantics on top.
DeepSeek Harness plugin: turn UI screenshots into structured, implementation-grade web frontend specs. Deterministic geometry (sharp) + optional vision-model semantics, merged into one JSON + Markdown spec.
Why it matters for design
- Pixel coordinates, spacing scales and token palettes are computed deterministically, not guessed by a vision model.
- Spec fields map straight to implementation: suggested tokens to design tokens, the spacing scale to CSS.
- Semantic roles supply intent while geometry supplies placement, merged into one JSON + Markdown spec.
What it covers
- Why this exists
- Configure the semantic layer
- The tool
- How to use the spec
- Status & caveats
How to install
dsh-ui-spec
-
Run this in a terminal.
dsh plugin --profile web add dsh-ui-spec -
Restart dsh web so it picks up the plugin.
-
Then describe the design you want. The harness picks up the plugin’s tools.
How to run it with DeepSeek Harness
-
Add the plugin; the geometry layer works offline with zero configuration.
-
Optionally point the semantic layer at any OpenAI-compatible vision endpoint.
-
Hand the agent a screenshot and build from the returned spec instead of the image.
Every plugin here is free to install and links to its real upstream source.
More in the DeepSeek Harness repo ↗More DeepSeek Harness design plugins
Vision & Input What it does modlens Gives text-only DeepSeek models plug-in vision: paste a screenshot, get structured evidence. 
Vision & Input What it does dsh-vision-toolkit Ten vision tools for UI restoration, grounding and pixel-diff verification. 
Vision & Input What it does dsh-media-skills Free eyes and a free paintbrush: paste-image reading plus watermark-free image generation. 
Canvas & Generative UI What it does dsh-openpencil The agent designs on a real, editable vector canvas instead of returning static images. Next step
Design with NEXGERCORE, without the setup
Install this plugin yourself, or run a whole curated design layer around DeepSeek Harness with NEXGERCORE. Bring your own key, own your output.