Add compact metadata fetch input

[mbret]

Summary

• replace the archive-reader-shaped metadata query with a fetcher-owned FetchMetadataInput
• add metadataInputFromResolvedArchive as the explicit archive-reader adapter
• update providers, search-term detection, scoring, tests, the HTTP API, playground, and documentation to use the compact input
• keep provider candidates as rich ResolvedMetadata results and bump the fetched entity schema to version 4

Why

The previous input accepted ResolvedMetadata or anything carrying it, which made archive-reader's rich result entity double as the metadata-fetcher's query model. Providers only understood a small subset of those fields, while the public API implied that covers, descriptions, subjects, layout metadata, and format-specific values were meaningful lookup inputs.

This change makes the intended asymmetry explicit: a small lookup-and-matching input goes into providers, and rich catalog metadata comes back.

Breaking changes

This is a clean API break and requires a semver-major release:

• fetchMetadata and MetadataProvider.search now accept only FetchMetadataInput
• scoreMetadataCandidate accepts FetchMetadataInput plus a rich candidate
• match signal fields use authors, publisher, and publishedYear
• fetched entity version is now 4
• projectGutenbergLookupFromMetadata is renamed to projectGutenbergLookupFromInput
• the HTTP API accepts the compact POST body and flat publisher / publishedYear query parameters

Validation

• npm run tsc — all 8 TypeScript projects passed
• npm test --workspace @prose-reader/metadata-fetcher — 98 tests passed
• npm test --workspace prose-reader-metadata-fetcher-api — 32 tests passed
• npm run build --workspace @prose-reader/metadata-fetcher — production build passed

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
prose-reader-demo	Ready	Preview	Aug 2, 2026 11:14am
prose-reader-front	Ready	Preview	Aug 2, 2026 11:14am