// the take
Inkline started as a Vue 3 component/design-system library, but the actual repo (changesets, skills, compiler tooling) shows it's grown into a cross-framework UI compiler that emits components for React, Angular, Svelte, Solid, Qwik and Astro from one authoring format, not just Vue. It's aimed at teams maintaining one design system across multiple frontend stacks, which is a much bigger ambition than the public description suggests.
The compiler does real per-framework lowering, not just transpilation — changesets like 'react-component-ref-forwardref', 'vue-jsx-named-slot-fill', and 'svelte-render-snippets' show it handling ref forwarding, slot/children mapping, and two-way binding idiomatically per target rather than with a lowest-common-denominator shim. There's a real CLI with typecheck gating, diagnostic source frames, and configurable report levels, which is the kind of DX investment that separates a usable compiler from a research project. Accessibility and a headless/styled component split are baked into the primitives from the start (e.g. 'angular-headless-styled-collapse'), which is far harder to retrofit than to design in up front.
The project description is badly out of date — it still says 'Vue.js 3 Design Systems' while most of the recent work is multi-framework compiler internals, so anyone evaluating it from the README/tagline alone will miss what it actually does now. A compiler that has to get ref handling, two-way binding, slot semantics, and JSX attribute casing right across six+ frameworks has an enormous edge-case surface, and that's exactly where subtle production bugs hide. The changeset history is dominated by granular fixes to core binding/slot/ref behavior, which reads as a compiler still stabilizing rather than one with settled semantics — treat anything beyond the original Vue 3 components as pre-1.0. At 1,470 stars and 76 forks, the community is small relative to the scope being taken on; there's little visible evidence of outside contributors who could carry the compiler internals if the core team slows down.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →