Releases: axtonliu/smart-illustrator
Releases · axtonliu/smart-illustrator
v1.4.0 - Tri-Engine System & Article Illustration Constraints
What's New
Tri-Engine System
- New Excalidraw engine for hand-drawn concept diagrams, comparisons, and simple flows
- Three-tier priority: Gemini → Excalidraw → Mermaid
- All diagram engines now output PNG by default
--engineparameter:auto/gemini/excalidraw/mermaid--mermaid-embedto fall back to code block embedding
Mermaid Usage Threshold Raised
- Mermaid restricted to complex diagrams only (>8 nodes, multi-layer architecture, multi-actor sequence diagrams)
- Simple concepts, comparisons, and basic flows now default to Excalidraw
Article Illustration Constraints
- Default aspect ratio changed from 3:4 to 16:9 for inline article images
- 3:4 and 1:1 prohibited for article illustrations (3:4 reserved for vertical platforms like Xiaohongshu)
- Whitespace ≥ 40%, single metaphor per image, visual echo with cover
Cover Safe Zone
- Core visuals must stay within center 71% height for 16:9 → 5:2 cross-platform compatibility
- Supports Newsletter, WeChat (16:9) and X Article (5:2 crop) from a single image
Dependencies (optional)
cd ~/.claude/skills/smart-illustrator/scripts && npm install && npx playwright install firefoxv1.3.0 - Mermaid Visual Enhancement
What's New
Mermaid Semantic Color Palette
- 7-color semantic palette with fixed meaning: input (green), process (purple), decision (red), action (orange), output (cyan), storage (yellow), meta (blue)
classDef+classsyntax examples for easy adoption- Arrow differentiation:
-->primary /-.->optional /==>emphasis - Layout rules: subgraph grouping, node limits (≤15 per diagram)
Cover Image Quality
- Ban decorative quotes (
「」,《》) around text in AI-generated covers
Documentation
- Dual-engine architecture diagram in README
- Background section with Make workflow video
- Improved SKILL.md with clipboard output and Mermaid rules
Full Changelog
v1.0.0 - Initial Public Release
🎉 Initial Public Release
Smart Illustrator - AI-powered article illustrations with intelligent position detection and cover learning system.
✨ Core Features
- Dual-Engine System: Auto-selects Mermaid (structured diagrams) or Gemini (creative visuals)
- Smart Position Detection: Analyzes article structure to identify optimal illustration points
- Cover Learning System: Learn from high-performing YouTube thumbnails
- Multi-Platform Support: YouTube, WeChat, Twitter, Xiaohongshu presets
- Extensible Styles: Light, Dark, Minimal, Cover, and custom styles
📦 What's Included
- Article illustration mode with auto position detection
- YouTube cover generation with best practices
- PPT/Slides batch generation
- Cover learning system (Gemini Vision analysis)
- Prompt management system (externalized prompts in
prompts/) - Comprehensive testing suite (11 test cases)
🚀 Quick Start
# Install
git clone https://github.com/axtonliu/smart-illustrator.git ~/.claude/skills/smart-illustrator
# Use
/smart-illustrator path/to/article.md📚 Documentation
- English: README.md
- 简体中文: README.zh-CN.md
⚠️ Status
Experimental - Best-effort maintenance. This is a public prototype that works for demos but may not cover all edge cases. PRs welcome!
🙏 Credits
Cover design best practices based on research of top YouTubers (Fireship, Theo, MKBHD). See references/cover-best-practices.md for details.