Ground truth for mixed.pdf (YYSI synthetic demo, 2026-11-11).
Pages 1 and 3 have a text layer; page 2 is a photo of a receipt (no text layer).

=== Page 1 (text layer) ===
Equipment handover file (demo)
This demo file has three pages. Pages 1 and 3 carry a text layer;
page 2 is a scanned image without any text layer.
Use this file to practise mixed-document handling.

=== Page 2 (image only — visible content, needs OCR) ===
Workshop receipt (demo)
Brake pads 1 x 38.00 38.00
Labour 1 x 45.00 45.00
Total 83.00
Paid in cash. Thank you!

=== Page 3 (text layer) ===
交接文件说明（演示）
第 2 页是扫描图片，没有文字层；第 1、3 页有文字层。
遇到这类混合文件时，逐页判断：有文字层的页直接提取，
扫描页先导出图片，再交给图片识字工具逐页处理。