CLI Reference

The EdgeSVG CLI is a single binary that can be installed via Cargo.

cargo install edgesvg

Global Options

edgesvg [OPTIONS] <COMMAND>

Option	Short	Description
--help	-h	Print help
--version	-V	Print version
--verbose	-v	Verbose output

convert

Convert a single image to SVG.

edgesvg convert [OPTIONS] <INPUT> [OUTPUT]

Arguments

Argument	Description
<INPUT>	Path to input image (PNG, JPG, WebP, GIF)
[OUTPUT]	Output SVG path (defaults to <INPUT>.svg)

Options

Option	Short	Default	Description
--method <METHOD>	-m	auto	Vectorization method: auto, logo, premium, smart, convert
--target-ssim <FLOAT>		0.995	Target SSIM for smart method
--max-file-size <BYTES>		none	Max output file size
--json	-j	false	Print quality report as JSON
--verbose	-v	false	Verbose output

Examples

# Basic conversion
edgesvg convert logo.png logo.svg

# Auto mode with JSON report
edgesvg convert input.png output.svg --method auto --json

# Target quality with smart mode
edgesvg convert input.png output.svg --method smart --target-ssim 0.998

# Premium mode, verbose
edgesvg convert illustration.png --method premium --verbose

Example JSON output

{
  "svg": "...",
  "requested_method": "Auto",
  "effective_method": "Logo",
  "fallback_from": null,
  "decision": "Image classified as logo (low unique colors, high edge density)",
  "report": {
    "score": 0.984,
    "grade": "A",
    "summary": "Excellent vectorization quality",
    "metrics": {
      "ssim": 0.9980,
      "edge_f1": 0.9870,
      "path_count": 4,
      "file_size": 1842
    }
  }
}

---

batch

Convert multiple images with glob patterns.

edgesvg batch [OPTIONS] <INPUT_GLOB> [OUTPUT_DIR]

Options

Option	Default	Description
--method	auto	Vectorization method for all inputs
--jobs	CPU count	Parallel worker threads
--json	false	Write report.json to output dir
--target-ssim	0.995	Quality target for smart method
--continue-on-error	false	Skip failing items instead of stopping

Examples

# Batch convert folder
edgesvg batch "assets/icons/*.png" assets/icons/svg/

# Parallel with 8 threads
edgesvg batch "logos/**/*.png" out/ --jobs 8

# Write report
edgesvg batch "*.png" svg/ --json

---

benchmark

Run quality benchmarks against a set of golden images.

edgesvg benchmark [OPTIONS] <GOLDEN_DIR>

Options

Option	Default	Description
--output	stdout	Output path for results CSV/JSON
--format	table	table, json, csv

Example

edgesvg benchmark golden_data/ --format json --output benchmark_results.json

---

inspect

Print quality metrics for an existing SVG against its source image.

edgesvg inspect <ORIGINAL_IMAGE> <SVG_FILE>

Example

edgesvg inspect logo.png logo.svg
# SSIM: 0.9980  EdgeF1: 0.9870  Grade: A

---

version

Print the installed version.

edgesvg version
# edgesvg 0.1.0 (Rust 1.82, built 2025-01-01)

---

Exit Codes

Code	Meaning
0	Success
1	Conversion failed
2	Quality threshold not met (smart mode)
3	Invalid arguments
4	I/O error

---

Environment Variables

Variable	Description
EDGESVG_LOG	Logging level: error, warn, info, debug, trace
EDGESVG_JOBS	Override default parallelism for batch commands
EDGESVG_TARGET_SSIM	Override default SSIM target