Vision & Input
dsh-vision-toolkit
Ten vision tools for UI restoration, grounding and pixel-diff verification.
What it is
A DeepSeek Harness-native bundle of ten vision tools: image Q&A, grounding and detection with pixel coordinates, long-screenshot OCR, cropping, color extraction, HTML screenshots and pixel diff. Tools mount progressively through a vision-tools skill.
DeepSeek Harness-native integration for agent-vision-toolkit: image Q&A, long-screenshot OCR, UI restoration, grounding, pixel diff, Artifacts, and Web UI.
Why it matters for design
- UI restoration closes the loop with numbers: a checked-in workflow iterates a rebuild from 6.04% pixel difference down to 0%.
- Grounding and detection return original-image pixel boxes, so the agent acts on coordinates instead of parsing prose.
- Infographics and hand-drawn sketches become editable HTML/CSS interfaces.
What it covers
- Why this exists
- Proven use cases
- Reference-to-pixel verification
- Tools
- Security and execution model
How to install
dsh-vision-toolkit
-
Run this in a terminal.
dsh plugin --profile web add @dsh-external/dsh-vision-toolkit -
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 to your web or headless profile and set a vision credential for the remote tools.
-
Activate the toolkit; the vision-tools skill mounts all ten tool schemas.
-
Rebuild a reference, then verify with vision_html_screenshot and vision_pixel_diff.
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-ui-spec Turns UI screenshots into implementation-grade specs: tokens, spacing scale, layout grid. 
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.