Hi everyone! I built Danqing (丹青), an open-source web studio, Python library, and Antigravity MCP plugin that helps researchers recolor finished scientific vector figures (.pdf / .svg from ggplot2, matplotlib/scanpy, GraphPad Prism, or Illustrator) in seconds—without re-running plotting scripts or altering a single coordinate, text glyph, or raw experimental image pixel (Western blots, H&E, microscopy).
Live Studio: https://danqing.onrender.com
GitHub (Apache-2.0): GitHub - owehann/danqing: New color for discovery, faithful to every datum — lossless PDF/SVG scientific figure recoloring studio & 277 verified palettes · GitHub
How Gemini 2.5 Flash is used
In Manuscript Match, researchers can type a target journal or study context (e.g., Nature Immunology, NEJM cohort, or single-cell spatial atlas). Gemini 2.5 Flash interprets the scholarly tone and domain constraints and pairs with a deterministic OKLCH perceptual comfort ranker over 3,434 verified palettes (including colorblind-safe, CNS/clinical, and homologous sequential/diverging ramps) to recommend 3 harmonious figure-set palettes—then exports ready-to-paste ggplot2 and matplotlib code.
Key Highlights
- Lossless Vector Recoloring & Evidence Shield: Modifies only vector fills/strokes while keeping embedded raster images bit-identical.
- Cross-Figure Consistency: Enforces
same_nameentity mapping across multi-panel or multi-figure manuscripts. - Agent-Native (
danqing_mcp): Ships as an Antigravity plugin + MCP server (danqing_inspect_figure,danqing_match_manuscript_palettes,danqing_recolor_figure,danqing_list_palettes).
Feedback from the Gemini & Antigravity developer community is very welcome!