Partner logos as bare tappable images, and loading skeletons for Docs/Videos

[simonhamp]

Two related polish passes on the Home, Docs and Videos tabs.

Agency partners: bare tappable logos

The partner cards laid out badly on iOS, and two of the three had no logo at all — they fell back to a lettered monogram because the native <image> renderer can't draw SVG and only Synergi published a raster.

• Rasterised all three logos from the official SVGs into transparent-background PNGs, with a light and dark variant each, bundled under public/img/partners. Nexcalia and Web Mavens come from the inline SVGs on nativephp.com/consulting; Synergi is regenerated from the SVGs already in the repo, whose previous PNGs had a baked-in white background.
• Variants are picked server-side via isDark(), so an appearance flip needs a re-render — hence the new AppearanceChanged listener on Home.
• Dropped the cards, taglines and chrome: at logo size the text was noise, so the section is now a centered column of bare tappable logos, each sized to its own aspect ratio. That removes the iOS layout issue along with the layout.

Bundled local paths work on both platforms (UIImage(contentsOfFile:) on iOS, Coil on Android), which is what let the rasters replace the remote-only URLs.

Docs / Videos: loading skeletons, only when needed

Both screens block in mount() on a network fetch, so a fresh push showed nothing until it returned. Both are now #[Lazy] with a placeholder() skeleton mirroring the real layout — Videos: featured thumb, section header, list rows; Docs: the TOC with its first section expanded — so the frame doesn't shift when content lands. Bones use theme tokens (works in both appearances) and are static, since the renderer has no pulse/shimmer support.

A skeleton only appears when there's actually something to wait for:

• Videos uses Cache::remember, so a warm cache returns within the frame — publishPlaceholder() bails on Cache::has() and paints the real feed.
• Docs needed more: DocsIndex::sections() is deliberately network-first, so a warm cache did not make mount fast, and skipping the skeleton on "cache exists" alone would have left the previous screen on display. DocsIndex::fresh() now returns the cached corpus only inside a short window, tracked by a separate marker key so the corpus itself still survives its full 24h as the offline fallback. mount() and publishPlaceholder() branch on the same condition, so they can't disagree.

The window is deliberately short: past it, sections() refetches (and the skeleton covers it), so a docs update lands without waiting out the 24h fallback TTL. Pull-to-refresh still forces a fetch anytime. FRESH_MINUTES is the single knob if you'd rather trade freshness for never seeing the skeleton.

Also removes the $loading prop and its branches from both screens — mount() always ran to completion before the first render, so those "Loading…" states were dead code that #[Lazy] now genuinely fills.

Also

Bumps nativephp/mobile to latest dev-main (c53cae8 → 1b61319) to pick up the merged stale-tab-render work. The #[Lazy] / placeholder() API the skeletons rely on is unchanged there.

Testing

• php artisan view:cache compiles all views, including the two new skeletons.
• php artisan test: 24 passed, 161 assertions. The one failure (ExampleTest, "No application encryption key") is pre-existing — this checkout has an empty .env — and unrelated.
• Verified the cache gating directly against a live cache store: cold → skeleton on both screens; warm → no skeleton on either; stale docs → skeleton plus refetch, with the 24h corpus intact as the offline fallback.
• Pint passes on every file touched. (app/Providers/NativeServiceProvider.php and app/NativeComponents/Concerns/InteractsWithDiscovery.php fail Pint, but both already do so on main.)

Worth a visual check on both platforms that the logo display sizes feel balanced, and that the skeletons line up with the loaded content.

🤖 Generated with Claude Code