Changelog

All notable changes to EdgeSVG are documented here. EdgeSVG follows Semantic Versioning.

---

Unreleased

• WASM SIMD acceleration for browsers that support it
• Streaming batch API
• SVG path simplification post-processing

---

0.1.0 — 2025-01-14

Added

Core engine

• Raster-to-vector engine with 5 methods: auto, logo, premium, smart, convert
• 16 quality metrics: SSIM, perceptual SSIM, edge similarity, edge precision, edge recall, EdgeF1, foreground IoU, color similarity, fidelity score, ΔE, topology score, PSNR, MAE, file size, path count, weighted path count
• Composite quality grade: A+, A, B, C, D, F
• Auto classification: detects logos, icons, illustrations, photos
• Smart mode: iterative quality search with configurable SSIM target
• Fallback chain: graceful degradation if preferred method fails

CLI

• edgesvg convert — single image conversion with --json report
• edgesvg batch — parallel batch processing with glob patterns
• edgesvg benchmark — golden-data benchmarking
• edgesvg inspect — quality check existing SVG vs. source
• All commands support --json output for CI integration

Python SDK (edgesvg)

• vectorize(path, **opts) — file-based conversion
• vectorize_bytes(data, **opts) — in-memory conversion
• PEP 561 type stubs
• Wheels for Python 3.9–3.13 on Linux, macOS, Windows

Node.js SDK (edgesvg-node)

• vectorize(path, opts?) — async file-based conversion
• vectorizeBuffer(buf, opts?) — async in-memory conversion
• Full TypeScript types
• Pre-built bindings for Node.js 18, 20, 22 on Linux, macOS, Windows

WebAssembly SDK (edgesvg-wasm)

• vectorize(bytes, opts?) — synchronous conversion in browser/edge
• Works with Vite, Webpack 5, Next.js
• Web Worker example for non-blocking UI
• CDN distribution via jsDelivr

Documentation

• Full API reference for all four SDKs
• Getting started guides for each platform
• Algorithm deep-dive
• All 16 quality metrics documented
• Vectorization modes explained
• Batch conversion guide
• CI/CD integration guide
• Interactive browser demo

---