PixelRAG-inspired visual research skill. Captures web pages as screenshots, extracts rich visual descriptions via vision analysis, and stores them in a searchable corpus. When researching new topics, searches past visual captures and re-examines matching screenshots with the new question — catching structure (tables, charts, layout, diagrams) that text-only parsing misses. The core insight: text-based RAG loses what's on screen; visual RAG preserves it.
PixelRAG-inspired visual research skill. Captures web pages as screenshots, extracts rich visual descriptions via vision analysis, and stores them in a searchable corpus. When researching new topics, searches past visual captures and re-examines matching screenshots with the new question — catching structure (tables, charts, layout, diagrams) that text-only parsing misses. The core insight: text-based RAG loses what's on screen; visual RAG preserves it.
# Visual RAG Research
## When to use this
- **Researching a topic** where page layout, tables, charts, diagrams, or visual structure matters
- **Comparing visual information** across multiple sources (competitor sites, pricing tables, feature comparisons)
- **Investigating a UI-heavy product** — dashboards, design tools, no-code platforms
- **Extracting data from visually-structured pages** where HTML parsing is lossy
- **Building a personal research corpus** that compounds across sessions
- A