ADR 0013: Ship monochrome tracing first before multi-color tracing

Decision: Ship monochrome-only tracing for v1; defer multi-colour. Why: All real dogfooding work was monochrome; adding colour would require unvalidated UX decisions (palette size, layer ordering, threshold semantics per layer) that couldn’t be closed before the release deadline.

Context

I want to support multi-colour / palette-based vectorisation eventually. The core algorithm (Potrace) and the current UI are built around a single threshold and a single fill colour. Adding multi-colour tracing would require palette detection, multiple passes or regions, and a different settings model.

The question this ADR answers is not “is monochrome enough forever?”, it isn’t. The question is “what should v1 ship?”, and that is a scope decision, made under a release deadline, informed by what I had actually been using the app for and what I could validate before publishing.

Options considered

Decision

Ship and iterate on monochrome tracing first. The first release focuses on single fill colour, threshold-based conversion with the existing parameters (threshold, corner policy, artifact area, smoothness, path optimisation). Multi-colour tracing is explicitly deferred to a later version, and the implementation path for it is the natural next step from ADR 0003: introduce vtracer alongside Potrace, route per-image based on a colour-mode flag, share the Sharp preprocessing layer.

Rationale

Why monochrome covers the use cases that mattered for v1. This is grounded in dogfooding, not market research. The work I had been doing with Vektory while building it, logos for my own project, icons for a friend’s design work, the assets for this very docs site (the Vektory logo in the taskbar, see ADR 0010), was almost entirely monochrome: logos and icons are typically one or two colours, designed flat, with colour applied downstream as a CSS or SVG fill attribute. That is the real workflow vector tools serve: take a raster mark, get a clean path, colour it later. Full-colour illustration vectorisation is a different category of need, served by different tools, and not where I or the people I had been making the app for were working.

Consequences

Positive

Negative

Follow-up