- YouTube: keep transcript segment classes and timestamps, so Obsidian Web Clipper's reader can follow the video, scroll and seek again
- YouTube: ignore braces inside strings when reading inline JSON
- Mastodon: don't merge replies into the post when handles are missing
- Footnotes: keep every footnote when a page mixes styles
- Markdown: keep task list checkboxes, and don't rewrite code blocks during cleanup
- C2 wiki: open every missing list level at once
- C2 wiki, X: time out stalled fetches after 4 seconds, so
parseAsync()can't hang - Parse: remove images from the output instead of the live page, skip retries for site extractors, and keep the cover image upgrade out of cached metadata
- Parse: don't change the page being parsed.
<noscript>images,srcSetattributes, and Medium, NYT and Substack content are now prepared on Defuddle's copy - Parse: leave Web Clipper's stamped shadow DOM on the page, so later parses can still read it