A page recipe is an assembly instruction, not an archive. It records which physical page to copy from which source, in what order. Rebuilding still requires the same local PDFs and permission to use them.
The entire grammar
A:1 one page A:1-3 ascending, inclusive: 1, 2, 3 A:5-3 descending, inclusive: 5, 4, 3 A:1,A:1 intentional repeated page A:2,B:1 pages from two sources
Aliases use uppercase letters and are assigned when files are selected. Spaces around separators are allowed. A comma separates items. Page zero, a page beyond the source count, an unknown alias or an empty item blocks export. Output is limited to 10,000 pages as a recipe-size ceiling, not a promise that your browser has enough memory for that many large pages.
Metadata is not identity
Saved JSON records version 1, the recipe, source aliases, filenames and page counts. A reload requires those names and counts to match. Two different PDFs can share both, so the check does not establish file identity. Verify versions and visible content. The JSON does not embed the PDFs; it can reveal filenames, so review it before sharing.
Page copying is deliberately narrower than editing
The implementation uses pdf-lib to copy selected pages into a newly created PDF. The source bytes are not modified. The browser offers a new download; do not choose an existing original’s filename when saving manually.
Do not rely on preservation of document-level bookmarks, attachments, form behavior, page labels, metadata or digital signatures. A new PDF is not the signed original. Encrypted PDFs are outside the supported scope, and the tool is not a password-removal service. It also does not OCR, redact, repair or validate PDF compliance.
A short check after export
- Confirm the output page count and visible order against the preview.
- Look at a few pages from each source for rendering or missing-content issues.
- Keep source PDFs and the recipe separate from the reading copy.
- If an interactive or legally significant feature matters, use a suitable specialist workflow rather than assuming a page copy preserved it.
All page processing occurs in the browser after public site assets load. Dependencies are bundled locally; no runtime CDN or document upload is used. For very large PDFs, memory use depends on the browser and device.