A Multi-View Dashboard with LLM-Guided Support for Astronomical Discovery
PreVue is a proof-of-concept astronomy interface designed to reduce fragmentation during early exploratory analysis. It combines navigable sky context, synchronized Pan-STARRS 2 grizy image views, catalog previews, derived photometry and SED-style summaries, and LLM-guided generation of additional comparison windows in one workspace.
The figure above illustrates the core workflow: users select a sky region, compare synchronized multi-band views and derived summaries, inspect magnified regions, and request additional comparison views through natural-language prompts. (A) Users select a sky region. (B--C) Linked panels synchronize the selection across multi-band views and derived summaries. (D) Magnified views support closer inspection. (E--F) Natural-language prompts generate additional dashboard views for side-by-side comparison.
- Region selection using celestial coordinates
- Synchronized crosshairs across multi-band
grizypanels - Magnified views for closer inspection
- Catalog previews and derived summaries
- LLM-guided generation of bounded comparison windows
- Coordinated multi-view exploration within a single workspace
git clone https://github.com/BJWOODS/PreVue.space.git
cd PreVue.space
npm install
npm run devFor LLM-assisted view generation, create a local .env file containing:
VITE_OPENAI_API_KEY=your_api_key_hereDo not expose a browser-side API key in a public deployment. Use a server-side proxy or backend endpoint for production use.
PreVue is an exploratory research prototype. Its image-derived summaries are intended for interaction-level previewing rather than calibrated photometric measurement or autonomous scientific interpretation.