docsor1212/academic-figures

> Publication-ready scientific figures from data — one command, verified output.
> 从数据一键生成投稿级论文图表：期刊合规、统计正确、中文零配置、纯本地运行。
39 chart types (bar/slope/volcano/upset set-intersections/oncology waterfall/raincloud/clinical swimmer/ridgeline joyplot/dumbbell gap chart/model calibration curve/decision curve analysis/clinical nomogram/restricted cubic spline/subgroup forest/competing-risk CIF/box/violin/
scatter/line/heatmap/clustered-heatmap/forest/KM/ROC/venn 4-set ellipse/Bland-Altman/PCA/
funnel/Galbraith radial/paired/dual-axis/stacked/composite/diagram/PRISMA 2020 …) · 9 journal presets (Nature/Lancet/Science/Cell/NEJM/JAMA/IEEE/CMA/
中文核心) · 9 color themes incl. colorblind-safe Okabe-Ito · 600dpi TIFF/PDF/SVG/EPS ·
fully local, zero telemetry.
## Quick Start
bash
# one-command figure (auto-picks the chart type)
python3 scripts/genfigure.py --quick -d data.csv
# Nature double-column submission (width/font/DPI per author guidelines)
python3 scripts/genfigure.py -t bar -d data.json -o fig.pdf --journal nature --column double --verify
# Cox multi-variable regression forest (raw per-subject data; Efron + PH test)
python3 scripts/genfigure.py -t forest -d patient.csv --stats cox --cox-time time --cox-event event --cox-cols "age,sex,treat"
# Chinese zero-config (auto CJK font detection)
python3 scripts/genfigure.py -t km -d survival.json -o km.png --cjk
## Why academic-figures
Journal compliance built-in: column widths, font sizes and DPI from official author
guidelines; --verify pixel-level text-overlap gate; auditpdf.py font-size gate —
the hard desk-reject reasons are guarded mechanically.
Statistical correctness: Cox (Efron ties + PH test), KM (risk table, log-rank,
median survival 95% CI), ROC (DeLong), metafunnel (Egger) — all validated against
reference implementations (lifelines) to machine precision.
Chinese zero-config: CJK font auto-detection, CMA/中文核心 presets, Chinese
diagnostics for every error (graded exit codes 0-6).
Safe by construction: fully local rendering, no network requests, no telemetry.
See SKILL.md "Safety & Data" statement.
## Docs
Full docs: SKILL.md (EN) · 中文文档见 SkillHub 页面
Getting started: references/quickstart.md · One-page cheatsheet:
references/cheatsheet.md · Limits & flag interactions: references/limits.md
Scenario templates: templates/ (39, each with a ready-to-run command)
Regression suite: tests/runtests.py (438 tests)
## Install
国内网络优先（无需访问 GitHub）：