@onadiet/pdf

The PDF format adapter for onadiet — put your PDFs on a diet, locally. Part of the @onadiet/* engine; you normally use the onadiet CLI (diet report.pdf --to 5mb), not this package directly.

It implements the @onadiet/core seams for PDFs:

• detect — is this a PDF? (header scan)
• weigh — where the bytes are (embedded images vs. everything else)
• slim — (v0.1 step 3) hit a size target by re-encoding embedded images

The work is done by permissive, best-in-class local tools: pdf-lib (MIT) for parse/rebuild and sharp/mozjpeg for the pixels. Output is standard PDF with JPEG (DCTDecode) images — the one lossy image filter valid inside a PDF (WebP/AVIF cannot live in one). No uploads, no copyleft engines bundled.

What's here (v0.1 step 2)

| Export | What it is | | ------------------------------- | -------------------------------------------------------------------- | | pdfAdapter | FormatAdapter — detect + weigh (kind 'pdf') | | sharpImageCodec | ImageCodec — decode/encode via sharp/mozjpeg (JPEG out) | | ssimMetric | QualityMetric — mean SSIM over 8×8 luma blocks (the quality floor) | | findImages / imageByteTotal | low-level pdf-lib image-XObject enumeration |

The extract→re-encode→re-embed capability these build on is proven by tests/integration/image-replace.probe.test.ts (a real PDF, measured smaller + still valid).

License

Apache-2.0.