Track estimated Anthropic API costs per request across all Claude API
call sites (recommender, analyze, artist-dive, generate-playlist, crate,
rabbit-hole, playlist-fix, timeline, compatibility). Log token usage and
estimated cost to the app logger. Aggregate costs in admin stats endpoint
and display total/today costs and token usage in the admin dashboard.
Add text/CSV export endpoints for playlists and saved recommendations.
Add export buttons to PlaylistDetail and Recommendations pages.
Add Open Graph and Twitter meta tags to index.html for better SEO.
- POST /api/recommendations/crate endpoint generates diverse crate of discoveries
- POST /api/recommendations/crate-save endpoint saves individual picks
- CrateDigger.tsx page with card UI, pass/save buttons, slide animations
- Progress bar, save counter, and end-of-crate stats summary
- Added to nav, routing, and API client
- Playlist Generator: describe a vibe, get a 15-30 song playlist, save or copy as text
- Artist Deep Dive: click any artist name for influences, best album, hidden gems, similar artists
- Music Timeline: visual decade breakdown of your taste with AI insight
- Nav updates: Create Playlist, Timeline links
Add mood_energy and mood_valence sliders that inject mood context into
AI recommendation prompts. Add "Surprise Me" button that generates
recommendations from a creative, unexpected angle without requiring
any user input. Includes backend endpoints, schema updates, and
full frontend UI integration.
- Add single and batch share endpoints with signed URL tokens
- Add public view endpoints (no auth required) for shared recommendations
- Add share button with clipboard copy to RecommendationCard
- Create SharedView page with Vynl branding and registration CTA
- Add /shared/:recId/:token public route in App.tsx
- Add shareRecommendation and getSharedRecommendation API functions
- Bandcamp: match artist+song, fall back to artist-only page
- YouTube fallback: if not on Bandcamp, link to YouTube music video search
- Fix My Playlist: AI analyzes playlist, finds outliers, suggests replacements
- User chooses recommendation count (5, 10, 15, 20)
