One Manifest for Responsive Images

The responsive-image cleanup started as a performance task: downscale assets, add srcSet/sizes, and improve LCP. The bigger problem turned out to be maintainability.

Image behavior was spread across scripts and components with repeated conventions:

• duplicated variant naming and path assumptions,
• multiple script aliases and legacy paths.

That made drift easy. A renamed variant or new profile could leave an old path assumption behind. I replaced that with a simpler pattern: generate one manifest at build time (src/generated/imageVariantManifest.json) and use shared helpers to resolve variants before rendering.

One manifest instead of scattered assumptions

yarn image:variants is now the canonical workflow. Lookup helpers read the generated metadata and pass resolved image paths and srcSet values into the shared ResponsiveImage component used for cover images. Inline Markdown images use the same manifest through the HTML conversion pipeline.

I also wanted to avoid manually creating every profile-specific variant when I added an image. For blog covers and inline Markdown images, the authoring path is now to add the source image, reference it in frontmatter or markdown, and run yarn image:variants.

Stricter failure modes

The manifest loader now rejects missing or empty required profile definitions (cover.card, cover.hero, inlineContent). Individual cover and inline variants can still fall back to the source image, so the hard failure protects the manifest schema rather than every cover or inline asset.