edf2csv 0.6.61 → 0.6.63

package/docs/CHANGELOG.md

@@ -3,6 +3,71 @@
 Notable changes to edf2csv. Versions follow [semantic versioning](https://semver.org); while the
 major version is 0, a minor bump may contain breaking changes.
 
+## 0.6.63
+
+### Fixed: phones could scroll the whole site sideways, and every heading faded in
+
+Three grids collapse to a single column on narrow screens, and all three collapsed to `1fr`.
+A plain `1fr` track has `min-width: auto`, so the one child with an unwrappable width — the
+scope readout's 420px minimum on the landing page, a CSV sample's longest line in the docs —
+set the track wider than the phone, and every element on the page shared one clipped right
+edge with 26px of sideways scroll to find it. The two-column desktop layouts already used
+`minmax(0, 1fr)` for exactly this reason; the single-column collapses now match them. The
+code blocks scroll inside their own boxes, which was always the intent.
+
+The scroll-reveal animation is gone. Every section's heading, lede and body used to fade
+upward as the reader reached it, which was defended in the README as arriving "in reading
+order" — but a page where every heading rises into place is indistinguishable from every
+other page where every heading rises into place, and the pattern costs something real here:
+the prerendered landing page shipped its text wrapped in inline `opacity: 0` until the bundle
+arrived to animate it. Content is now simply present. Motion stays where it argues something
+— the hero trace turning into rows, the 765 interpolated dots appearing out of nowhere — and
+in interaction feedback. The hero's own entrance became three lines of CSS, which paint with
+the HTML instead of after it.
+
+Two smaller things. The documentation grid on the landing page held eleven cards in a
+four-column grid, leaving a permanent empty cell; a twelfth card now points at
+llms-full.txt, the whole documentation as one plain-text file for pasting into a coding
+agent — a real destination this site already served without linking to it. And the footer
+finally links the npm package, which no page on the site did.
+
+## 0.6.62
+
+### Changed: the homepage is real HTML, and a pasted link unfurls into something
+
+Two findings from an SEO pass over the website, both the same shape: the site said less about
+itself to machines than it does to people.
+
+The documentation pages have been prerendered static HTML from the start, precisely so that
+crawlers that do not run JavaScript — which is most of the AI crawlers the docs were written
+for — get the full text. The landing page never got the same treatment: it was an empty
+`<div id="root">` until the bundle arrived, so the one page that explains what the tool is was
+blank to exactly the readers the prerendering exists for. It is now server-rendered at build
+time through Vite's SSR transform and hydrated in the browser, and the build fails if the
+rendered landing ever comes back as a stub, the same check the documentation pages have.
+
+A pasted link had no image at all: `twitter:card` said `summary` and no `og:image` existed, so
+Slack, iMessage and the rest unfurled the URL into two lines of grey text. There is now a
+1200x630 card in the site's own oscilloscope idiom — the H1, the command, a trace — referenced
+with `summary_large_image` from every page. One committed PNG rather than a per-page generator,
+because eleven near-identical title cards would say less than one good one.
+
+The smaller corrections follow the same principle. The sitemap dropped `<priority>`, which
+Google states it ignores, for `<lastmod>` read from each page's commit history — and a page
+whose date git cannot supply gets no tag, because a fabricated date teaches crawlers to
+distrust all of them. The title now leads with the phrase people type ("Convert EDF to CSV
+from the command line") while the H1 keeps the site's own voice. Documentation pages carry
+BreadcrumbList and a `dateModified` from the same commit history; the SoftwareApplication
+entity gained `sameAs` links tying the site, the repository and the npm package into one
+thing, and a `softwareVersion` read from package.json at build. It still carries no
+aggregateRating and no download count: the schema for star ratings is the one piece of SEO
+this site will not do, because inventing social proof is fabricating data on a site whose
+argument is that fabricating data is the problem.
+
+Also fixed on the way through: the landing page never inlined the theme-restore script the
+documentation pages have, so a reader who had chosen light mode got a flash of dark on every
+visit to the homepage.
+
 ## 0.6.61
 
 ### Added: the pyEDFlib cross-check runs weekly instead of never

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "edf2csv",
-  "version": "0.6.61",
+  "version": "0.6.63",
   "description": "Convert EDF, EDF+ and BDF biosignal recordings (European Data Format) to CSV from the command line. Local, streaming, and never resamples or alters units.",
   "keywords": [
     "edf",