versed-pdf 1.2.0

Semantic PDF-to-Markdown engine for Arabic and bilingual texts with local routing and repair

versed

Local PDF-to-Markdown tooling for Arabic and bilingual texts.

It repairs broken extraction, decodes QCF Quran fonts, classifies pages, and renders semantic Markdown from local PDFs. It also normalizes parsed OpenITI block JSON into the same public Document and markdown/plain-text outputs.

Install

pip install versed-pdf
pip install versed-pdf[pdf]
pip install versed-pdf[pdf,ocr]

Quick start

from versed import extract_document

result = extract_document("book.pdf", title="Book")
print(result.markdown)

OpenITI normalization:

from versed import build_openiti_markdown

parsed = {
    "metadata": {"title": "رسالة"},
    "content": [
        {"type": "title", "content": "رسالة"},
        {"type": "pageNumber", "content": {"volume": "01", "page": "218"}},
        {"type": "paragraph", "content": "قال المصنف ..."},
    ],
}

result = build_openiti_markdown(parsed)
print(result.markdown)

CLI

versed repair-text "taf߬l"
versed detect book.pdf
versed classify book.pdf
versed extract book.pdf -o book.md

Public modules

• versed.repair: Sabon mojibake repair helpers
• versed.qcf: QCF Quran font decoding
• versed.classify: local page classification and backend selection
• versed.routing: cost-aware routing heuristics
• versed.layout: aligned words to semantic blocks
• versed.markdown: semantic blocks to Markdown/plain text
• versed.openiti: parsed OpenITI block JSON to Document and markdown
• versed.extract: end-to-end local extraction

License

MIT