[isnogudus]

Adds GET /library/albums and GET /library/artists, listing the albums saved in the user's library and the artists they follow, most recent first, so a client can browse the whole library through the daemon:

• The lists come from the collection service added in feat: save to Liked Songs and add to playlists #403: saved albums share the collection set with Liked Songs, followed artists are the artist set.
• Names, artists, release year and images come from one batched extended-metadata request per 100 entities (ALBUM_V4 / ARTIST_V4), with image URLs built like album_cover_url (honouring image_size). Entries whose metadata does not resolve are still listed by URI.
• Both lists are cached per user for a minute and paged with offset/limit like /library/playlists; the paging validation is now shared by the three listings.
• Albums and artists can be opened or played by URI through the existing /context/tracks and /player/play.

Testing

• Unit tests for the newest-first ordering (skipping removed items and other types), the per-user cache with TTL, paging bounds and the album/artist mapping including missing metadata; API tests for routing and validation.
• Ran against a real account: 98 saved albums and 38 followed artists with names, years and cover URLs.
• go mod tidy, go generate ./..., gofmt, build and go test -tags "test_unit test_integration" ./... are clean.

🤖 Generated with Claude Code

[devgianlu]

Is the cache really needed? It adds a bunch of complexity

[isnogudus]

Is the cache really needed? It adds a bunch of complexity

No, it isn't. I added it because the collection service only pages by token, and saved albums share the "collection" set with Liked Songs, so every page meant re-reading the whole set and looking up every album. Looking up only the requested page fixes most of that, and on a real library (98 saved albums in a set of ~1200 items, 39 followed artists) a request now takes 0.1–0.3 s.

So I removed it: every request now reads the set and pages it in memory like /library/playlists does, and extended metadata is fetched only for the page's entities. That drops cachedList, its TTL and clock hook, and the two fields on App. Rebased onto current master as well.

[devgianlu]

BTW your git setup is broken in such a way it does not mark you as a contributor