doc2mark

doc2mark converts documents into Markdown and structured Python objects. It supports PDFs, modern Office files, images, text/data files, markup files, and legacy Office files through LibreOffice.

The default Python API is safe for text-only extraction without OCR credentials. OCR providers are initialized only when OCR is requested.

User Guide

• Quickstart
  • Install
  • The 30-second example
  • Output formats
  • Text-only vs OCR
  • Picking a provider
  • From the command line
  • Batch processing
  • Where to next
• Command Line
• Supported formats & how each is handled
  • How a format is detected
  • Extension → processor → OCR
  • Office formats (DOCX, PPTX, XLSX)
  • PDF
  • Legacy Office formats (DOC, PPT, PPS, XLS, RTF)
  • Images (PNG, JPG, JPEG, WEBP, TIFF, TIF, BMP, GIF, HEIC, HEIF, AVIF)
  • Text & data (TXT, CSV, TSV, JSON, JSONL)
  • Markup (HTML, XML, Markdown)
  • Email (EML)
  • Choosing whether OCR runs
• OCR
  • The OCR facade
  • What a result looks like
  • Tasks
  • Raw and legacy modes
  • Tables
  • Providers
  • Concurrency
  • Using OCR with the document loader
  • Deprecation notice
  • See also
• Content-aware OCR policy
  • The guiding principle
  • Layer 1 – the document strategy
  • Layer 2 – the Office image route
  • Layer 3 – the per-image job-router (task="auto")
  • Layer 4 – page_markdown synthesis and the coverage guard
  • Putting it together
• Complex Table Preservation
  • Why rule-based instead of OCR
  • The PDF path: geometric span detection
  • The Office path: OOXML grid spans
  • From TableData to clean HTML
  • Choosing the output style
  • Worked example: a merged-cell table
  • The OCR table path and Table.html
  • to_markdown() prefers html
  • Reading tables from a result
• Contextual OCR strategy
  • The idea
  • The context_pages tiers
  • How the window PDF is built
  • How each provider attaches the context PDF
  • It augments – it never replaces the verbatim layer
  • Enabling it
  • Cost and payload, honestly
• OCR Result Caching
  • Quick start
  • Factory helper
  • Using with UnifiedDocumentLoader
  • MemoryOCRCache
  • RedisOCRCache
  • Cache statistics

Reference

• API Reference
  • Convenience Functions
  • UnifiedDocumentLoader
  • OCR
  • Structured OCR Schema
  • OCR Caching
  • Chunking
  • Core Types & Exceptions
• Development
  • GitHub Pages