0.1.0

First release.

Converts PDF, Word, PowerPoint, Excel and image files into Markdown notes,
with their images. No Python, no pip, no cloud service — nothing you convert
leaves your machine.

Each format is read through its own structure rather than one generic
abstraction, which is what keeps the output clean:

• Word — heading levels from Word's own styles, ordered vs. bullet lists
  resolved through numbering.xml, tables, hyperlinks, bold/italic
• PowerPoint — one section per slide in presentation order, title
  placeholders as headings, speaker notes, slide tables
• Excel — dates instead of serial numbers, 27.38 instead of
  27.383982300884924, hidden helper sheets skipped
• PDF — multi-column layouts read in order, headings from font size,
  paragraphs rejoined and de-hyphenated, running headers and footers dropped
• Images — local OCR via Tesseract

Images are pulled out of every format and embedded where they sat — a figure
inside a Word table cell lands in that table cell. Identical images are
written once.

Installing

Copy main.js and manifest.json into
<vault>/.obsidian/plugins/convert-to-markdown/, then enable Convert to
Markdown under Community plugins.

Desktop only.

Note on OCR

Converting an image file downloads the recognition engine and English model
(~9 MB) on first use, then caches it. The image itself is never uploaded. On
machines where that CDN is blocked, point the OCR engine folder setting at
a vault folder holding the two files and OCR makes no network request at all.

Known limits

• Scanned PDFs are reported as an error rather than OCR'd
• Column detection is geometric, so a layout with no gutter running through it
  can still come out misordered
• Excel formulas export as their last-calculated value