dsh-vision-toolkit
DSH Vision Toolkit - Vision Tools for Text-Only Agents
Quick facts
- Best for
- DSH Vision Toolkit - Vision Tools for Text-Only Agents
- Pricing
- Free
- Editor rating
- 4.5 / 5
- Community saves
- 0
About dsh-vision-toolkit
DSH Vision Toolkit is an MIT-licensed DeepSeek Harness plugin that packages agent-vision-toolkit as native visual tools. Its README documents image Q&A, long-screenshot OCR, grounding, UI restoration, pixel verification, managed artifacts, web settings, a dsh plugin install command, Node.js and Python runtime requirements, and 168 verified tests.
Pros
- Intent-aware image question answering for text-only agents
- Long-screenshot OCR for extracting text from large screenshots
- Original-pixel grounding for coordinates, colors, geometry, and visual evidence
- UI restoration workflows that turn sketches or screenshots into editable interfaces
- Pixel diff verification for checking rendered output against references
- Managed artifacts and DeepSeek Harness Web presentation
- Native DSH plugin installation with structured schemas
Cons
Pricing
Open source
$0
- • Source code available on GitHub
- • No listed SaaS subscription required
- • Users pay only for their own compute, API, or hosting costs
