# Selecting pages & capping large PDFs

OCR and extraction cost scale with pages. You rarely need every page of a
long PDF, so the platform lets you select pages per document and cap large
PDFs across a whole batch.

## Single document

On **New document** (`/documents/new`) the **Pages** control offers:

- **All pages** — the default.
- **First N pages** — `1-N`.
- **First N pages + last page** — `1-N,last`. The `last` token resolves to
  the document's final page server-side, so you don't need to know the page
  count.
- **Custom** — a print-style spec like `1-3, 5, 12-14, last`.

The selection becomes the `page_range` form field on `POST /v1/documents`.
The PDF page limit applies to the **selected** count, so a range pulls a few
pages out of a very large PDF. Original page numbers are preserved on the
result (`canonical.pages[].source_page_number`).

## A whole batch

On **New batch** (`/batches/new`) **Page handling** applies to every file:

- **All pages.**
- **Cap large PDFs** — *first N pages + last page, for PDFs over M pages*.
  Each file is expanded individually from its own page count: a 400-page PDF
  becomes `1-N,last`; a 12-page PDF runs in full. This is the one-control way
  to keep a heterogeneous batch cheap — set it once and submit.
- **Same page range for every file** (uploads only) — one print-style spec
  applied uniformly.

`Cap large PDFs` sends a `page_cap_policy`
(`{over_pages, first_pages, include_last}`); the uniform range sends
`page_range`. They are **mutually exclusive**. The cap policy also applies to
source-pull batches — it is expanded as each remote file is discovered.

## The `page_range` DSL

`page_range` is comma-separated, 1-indexed, case-insensitive:

| Spec | Pages |
|---|---|
| `all` / empty / omitted | every page |
| `1-3,5,12-14` | 1,2,3,5,12,13,14 |
| `last` | the final page |
| `1-10,last` | first 10 + the last page |

A spec that lands entirely outside the document is a **400**; a spec that
partly overshoots is clamped, the way a printer would.