- Architecture overhaul: HTML→Markdown conversion now happens client-side in the browser instead of server-side in Python
- Content is now added as a note instead of a code cell that runs
url2note() - Refactored all code to follow fastai style guide
- Turndown.js integration for HTML→Markdown conversion
- GFM plugin for tables, strikethrough, task lists
- MathML→LaTeX conversion extracts TeX from
<annotation>elements - ToC extraction for arXiv and other sites with table of contents
- Smart content selection picks largest matching element to avoid grabbing comments
#aisuffix on images for solveit AI visibility- Element filtering removes nav, buttons, popups, cookie banners, etc.
scriptingpermission andweb_accessible_resourcesin manifest
- JS-rendered pages (SPAs, React apps) now work since we read from the live DOM
- Bracket escaping issue where
\[...\]triggered KaTeX block math
turndown.js(v7.2.2)turndown-plugin-gfm.js
- Basic URL import via
url2note()Python call