Release v1.3.0

Added

• Optional ML document-corner detector: pass detector: 'ml' to scanDocument or Scanner to use a neural corner detector — a channel-slimmed SimCC model (DocCornerNet, 456 K params, ~1.9 MB) paired with a custom minimal ONNX Runtime Web WASM build (~1.5 MB, 88 % smaller than stock ort-web). Classical users pay nothing: the detector is fully opt-in, lazy-loaded, and gated behind the optional peer dependency onnxruntime-web.
• scanic-ml companion package (npm install scanic-ml): the model (.ort format) and the custom ORT WASM loader. Published to npm and served from jsDelivr by default — no self-hosting required. Self-hosting supported via the ml.assetBaseUrl option.
• result.score: when using the ML detector, scanDocument returns a score field — the model's P(document present) confidence, 0–1.
• ml options namespace: assetBaseUrl, modelUrl, wasmPaths, modelBytes, numThreads, and minScore.

Installation

npm install scanic@1.3.0

Or via CDN:

<script src="https://unpkg.com/scanic@1.3.0/dist/scanic.js"></script>

What's Changed

• ci: bump github/codeql-action from 3 to 4 by @dependabot[bot] in #12
• ci: bump actions/setup-node from 5 to 6 by @dependabot[bot] in #13
• ci: bump actions/checkout from 5 to 7 by @dependabot[bot] in #14
• ci: bump softprops/action-gh-release from 2 to 3 by @dependabot[bot] in #15
• chore: bump the dev-dependencies group with 4 updates by @dependabot[bot] in #16

New Contributors

• @dependabot[bot] made their first contribution in #12

Full Changelog: v1.2.0...v1.3.0