v0.7.2

What's changed

This release improves Markdown output quality for PDFs without an embedded table of contents and fixes several extraction issues in multi-column and table-heavy documents.

Changed

• Set the release version baseline to v0.7.2.
• Single-file output no longer injects a generated top-level Table of Contents when the source PDF does not include an embedded TOC.
• Single-file conversion without an embedded PDF TOC now falls back to one document-level heading instead of noisy heuristic heading splits.
• Improved margin-noise normalization for short PDFs so recurring headers and footers are filtered more reliably.

Fixed

• Preserved reading order for columnar PDFs with mixed text and tables by processing text and tables per column instead of interleaving content by global vertical position.
• Fixed dropped text near table boundaries by filtering only significant line/table overlap.
• Fixed malformed row splitting that could corrupt Markdown table rendering.
• Prevented paragraph stitching across column resets, reducing out-of-order content in two-column documents.