[arjunpredev]

What broke

MarkdownImages (the extension that opens Markdown images as image cards) registers under the paragraph Markdown token so its parser runs first. @tiptap/markdown looks up a node's writer through the parse registry first too, so it asked MarkdownImages how to write paragraphs. It had no renderMarkdown, so every paragraph, list item and table cell saved as an empty string in the post's Markdown copy.

The newsletter's text/plain part is built from that copy (plainBody(post.markdown)), so a post saved from the editor mailed a text part with only headings, image alt text and empty bullets, next to a full HTML part.

Fix

• MarkdownImages.renderMarkdown writes paragraphs the way the paragraph extension does (children, and   for a second empty paragraph in a row).
• newsletterEmail builds the text part from the Markdown copy only when it holds at least 80% of the HTML body's words; otherwise from the HTML body. Posts saved before this fix send their full text without a re-save.
• plainBody writes HTML lists as - item lines and tables as one a | b line per row, with blank lines between blocks.

Checked

• Round trip through the editor's extensions (Markdown to document to Markdown): before, paragraphs, bullets and table cells came back empty; after, all text is kept.
• The text part for an affected post: 1.5 KB before (headings and alt text), 7.6 KB after with every paragraph, bullet and table row.
• bun run typecheck, admin build, leak guard.

🤖 Generated with Claude Code