What problem does it solve? Researchers often have data but don't know which chart best supports their argument, and generic plotting produces figures that fail journal requirements—wrong sizes, missing CJK glyphs, colorblind-hostile palettes, or mean-bar charts that hide small-sample distributions. ## Core Features & Use Cases - Advisor-first workflow: Runs exploratory data profiling (column types, sample sizes, skewness, outliers, correlations) before recommending chart types, and actively intercepts classic mistakes like mean bars for n<10 groups, dual Y-axes, pie charts, and rainbow colormaps. - Journal-spec rendering: Applies Nature / Science / IEEE / Elsevier / PNAS / Chinese-journal presets for column width, font size, DPI, and vector formats, with automatic CJK font configuration and minus-sign fixes. - Post-render self-check loop: Renders PNG previews, programmatically audits missing glyphs, text clipping, and tick overlap, then re-reads the image to verify legend occlusion and panel-label alignment before exporting PDF/SVG/PNG with grayscale previews. - Use Case: You drop in a results.csv and ask for a paper figure; the skill profiles the data, asks what claim the figure should support, recommends box + stripplot over a mean bar chart, renders it at Nature single-column size, and exports audited vector files. ## Quick Start Ask the assistant to analyze your CSV data file and create a publication-ready figure for your target journal, letting it recommend the chart type first.
