wtf is appnz-vectorizer?

appnz-vectorizer is a small tool that turns raster images, meaning ordinary pixel-based pictures like PNGs or JPGs, into SVG files, which are vector graphics that can be resized and edited without losing quality. It does the conversion using VTracer, a fast tracing engine written in Rust that can use multiple processor threads at once. The project is built as a Cog, a packaging format for running machine learning style tools in a container, and it is meant to sit alongside image generating AI models. The idea is that you can generate pixel art or images with any generative model you like, then hand the result to this tool for a clean, deterministic, production ready conversion into an editable vector format, rather than relying on the AI model itself to produce vectors directly. You can run it in a few ways: directly through the cog predict command with an input image and chosen preset, as a Docker container pulled from GitHub's container registry, or deployed to the app.nz platform as a hosted cog or app. There is also a Node.js test file for exercising the client that talks to the hosted queue system, which reports progress, handles errors and timeouts, and checks the shape of returned results. Four presets are offered for different kinds of source images: poster, logo, photo, and a hard edged pixel preset suited to pixel art. Input images are capped at 30 megapixels and path precision is limited so the resulting SVG files stay a reasonable size. Both the adapter code and the underlying VTracer tracing engine are MIT licensed, with further license details in a THIRD_PARTY file.