feat: hybrid table extraction with optional plumber/camelot backends

[FranciscoJBL]

This pull request adds experimental PDF table extraction support to MarkItDown, allowing users to extract tables from PDFs as markdown tables using optional dependencies. The feature is exposed via a new --pdf-tables CLI flag and a corresponding Python API parameter, with support for multiple extraction modes. Documentation and tests are included to explain usage and verify behavior.

PDF Table Extraction Feature

• Added experimental table extraction for PDFs, selectable via the new --pdf-tables CLI flag and pdf_tables Python API parameter. Supports four modes: none (default, plain text), plumber (uses pdfplumber), camelot (uses camelot), and auto (tries plumber then camelot, falls back to plain text). [1] [2] [3]

Documentation Updates

• Updated README.md to document the new table extraction feature, usage examples, supported modes, installation of optional dependencies, and caveats. [1] [2]

Dependency Management

• Added a new optional dependency group pdf-tables in pyproject.toml, including pdfminer.six, pdfplumber, and camelot-py.

Codebase Changes

• Modified the PDF converter to detect and use optional table extraction libraries, format extracted tables as markdown, and gracefully fall back if dependencies are missing. [1] [2] [3]
• Updated CLI and API argument handling to pass through the pdf_tables option to converters. [1] [2] [3]

Testing

• Added new tests for PDF table extraction covering different modes and fallback behavior, using in-memory PDF generation for robustness.

This PR address issue #1419

[FranciscoJBL]

@microsoft-github-policy-service agree