Skip to content

v1.1.1 — CMS-pattern preprocessing

Choose a tag to compare

View all tags
@syswave-dev syswave-dev released this 27 Apr 10:28
· 122 commits to main since this release
v1.1.1
be3ca87

What's fixed

Two small preprocessing steps in `cleanDom` that improve extraction on CMS-driven sites without affecting plain blog/article content.

1. Readonly `` values surface as ``. Click-to-copy widgets (API model slugs, embed snippets, share links) hide their content in `<input type="text" readonly value="…">`, which node-html-markdown silently drops as form chrome. They now come through as inline code. Only `readonly` text inputs are transformed; real forms stay untouched.

2. Bare-UUID alt-text on images is dropped. Strapi/Directus/Hygraph/etc. leak asset IDs as the `alt` attribute. A strict UUID-v4 regex on the alt — drop the alt only, keep the image link. Descriptive alt text is preserved.

Verified

Patched extraction was diffed against unmodified extraction across 9 representative URLs:

  • paulgraham.com, hillelwayne.com, 12factor.net, joelonsoftware.com, grugbrain.dev, RFC 2324, Wikipedia, Hacker News frontpage → byte-identical output
  • mistral.ai/pricing#api → expected diff: model slugs (`mistral-large-latest`, `devstral-medium-latest`, …) now surface as `code`; dozens of UUID alts dropped

Tests

209 / 209 green (4 new covering both transforms + their negative cases).

Assets 2
Loading