equinor/neqsim/paperlab-source-pdf-to-html

Skill: PaperLab Source PDF to HTML

Purpose

Convert or plan conversion of source PDFs into clean, searchable HTML for PaperLab books. Use this when lecture PDFs, public reports, exam PDFs, or standards excerpts need to be searchable by agents and linked to chapter figures, captions, and source manifests.

Inputs

• Book directory with source_manifest.json, or a --source-root folder.
• PDF files from lectures, exercises, exams, public reports, or approved

reference documents.

• Optional OCR output for scanned/image-only PDFs.

Workflow

1. Inventory source files with book-source-inventory.
2. Run python paperflow.py book-source-pdf-html-plan <book_dir> to create a

conversion manifest.

1. Convert PDFs using the selected local converter. A robust converter should:

• preserve the source PDF,
• extract text and images,
• fall back to OCR or PyMuPDF when table-heavy PDFs fail,
• write HTML under an _HTML/ folder next to the source category.

1. Feed converted HTML and extracted images into technical-figure-understanding

and figure_dossier.json.

Output

• source_pdf_html_plan.json
• source_pdf_html_plan.md
• Converted HTML files and extracted resources when the conversion script is run.

Quality Checks

• Count converted PDFs against the plan.
• Flag PDFs with very low extracted text length for OCR.
• Do not silently drop pages or figures.
• Record source path, output path, page count, text length, image count, and

extraction method when the converter supports it.

Safety Rules

• Never delete or overwrite source PDFs.
• Do not publish converted HTML for non-public or licensed material unless that

is explicitly allowed.

• For standards, extract only metadata or allowed excerpts unless the standard's

license permits broader reproduction.