Stacks.Books Works/editions Integration

US-1.1.1 marks a quiet revolution: every book upload now unfolds in two steps - identify first, confirm second - turning chaotic data into clarity. The shift isn’t just technical; it’s cultural, reshaping how US readers manage digital collections. The identify/2 call doesn’t commit yet, just surfaces candidates like a curated preview - relying on vision APIs and ISBN tying to prevent chaos. This two-step flow stops duplication before it starts, especially critical when TikTok trends turn obscure editions viral. Behind the scenes, Books.confirm/2 now checks for duplicates via ISBN or fuzzy title matching, and merge_edition/2 creates new works from conflicting versions - ensuring every shelf placement is intentional. Users see full edition details, including community read counts pulled from wh.mart_community_read_count, even if inventory’s still sparse. Platform-wide search now spans public shelves, listings, and partner stock, thanks to resolve_visibility/2. But here’s the catch: cover images - once deferred - now trigger real async validation via associate_isbn/4 in the AI layer, linking visuals to edition identities. The entire system faces no hidden traps, but safety demands attention: always verify cover association via HMAC, and never trust the client-side image. Is this workflow too slow? For many, the precision wins. Is the event flow solid? Every edit, merge, and search now breathes through Stacks.Events. The real test? Can we keep 84%+ accuracy across merged works without breaking the UX? That’s the bottom line - does your next upload feel like discovery, not noise?