@noidme/mac-print

Consent-native device identification. The browser signal probes that feed a device visitorId — canvas, WebGL (+ extensions), audio, ~64 fonts, WebGPU, media codecs, UA-CH, speech voices, screen, and math/libm. Zero-dependency, CSP-safe, and fail-closed by construction: a probe for a purpose that is not granted is never even constructed.

Part of the noidme SDK family: @noidme/consent provides the consent decision (the window.noidme bridge); mac-print collects the signals; the mac-print server engine turns them into a stable, tenant-scoped, epoch-rotated visitorId (with a drift-tolerant linker). For bot/fraud signals see @noidme/am-i-bot; for the all-in-one agent see @noidme/glassprint.

Install

npm i @noidme/mac-print

Or drop-in via a <script> (exposes window.MacPrint):

<script src="https://cdn.jsdelivr.net/npm/@noidme/mac-print/dist/mac-print.global.js"></script>

Use

Consent is read from window.noidme (installed by @noidme/consent). Collection is gated per purpose:

import { ConsentGate, resolveEngine, DEVICE_ID_PROBES } from '@noidme/mac-print';

const gate = new ConsentGate(await resolveEngine(2000)); // waits for the consent engine, fail-closed
const signals = [];
for (const factory of DEVICE_ID_PROBES) {
  if (!gate.granted(factory.purpose)) continue;   // never construct a probe for a denied purpose
  const s = await factory.create(gate).read();    // never throws; null when the API is absent
  if (s) signals.push(s);
}
// → seal `signals` and POST to your mac-print server, which returns the visitorId.

Each probe emits an opaque, already-canonicalized SignalValue ({ key, axis, value }). The raw, untrusted strings are only entropy inputs — identity + anti-spoof reasoning happens server-side.

Design

• Fail-closed / consent-native. No engine, or a denied purpose ⇒ nothing is collected.
• Never throws. Every probe returns null on absence/failure (e.g. no OffscreenCanvas in Node).
• CSP-safe. No external fonts/scripts/styles; uses OffscreenCanvas.measureText, not DOM injection.
• Entropy per axis. Probes declare an axis (gpu/cpu-fpu/os/display) so correlated signals are de-duplicated by the server before scoring.

License

MIT.