gongshichen/luminacode/pdf

Treat PDFs as visual artifacts. Prefer rendering pages to images and inspecting the result over relying only on extracted text.

Recommended workflow:

• Use pdftoppm or an equivalent renderer to convert pages to PNG when available.
• Use structured libraries such as pypdf, pdfplumber, or reportlab for extraction and generation instead of ad hoc byte edits.
• Keep temporary render outputs under tmp/pdfs/ and final artifacts under output/pdf/ unless the user gives another path.
• Re-render after meaningful edits and verify margins, clipping, tables, images, page numbers, headers, and footers.

If a dependency is missing, state the exact missing tool and the command needed to install it. Do not deliver a generated PDF as complete until the latest render has been visually checked or you have clearly explained why rendering was not possible.