DSH Vision Toolkit

Overview

A Harness-native integration for image Q&A, long-screenshot OCR, visual grounding, UI restoration, pixel comparison, Artifacts, and Web UI presentation.

What it does

• Intent-aware image questions and long-screenshot OCR
• UI restoration, grounding, and pixel-diff workflows
• Local tools plus configurable OpenAI-compatible vision endpoints

Required permissions

• May access configured vision endpoints and install a managed Python runtime; not yet directory-reviewed.