Recommend titles from taste profile with one-click watchlist

[ALeonard9]

Story

As a druthers user, I want the app to recommend titles based on my taste profile and let me add them to my watchlist in one click, so that I can discover new things to watch/read/play without leaving the recommendation surface.

Acceptance Criteria

• A recommendations surface exists (e.g. a section on the movies/TV/books/games home page, or a dedicated page) showing a list of suggested titles.
• Recommendations are derived from the user's existing taste signals (e.g. what they've ranked highly, marked completed, or watchlisted) rather than being generic/popularity-only.
• Each recommended title has a one-click action that adds it directly to the user's watchlist, without navigating to a search/detail page first.
• After adding, the item reflects on the user's actual watchlist (e.g. movies/watchlist) and the recommendation UI confirms the action (e.g. button state change).
• Recommended titles the user has already tracked (ranked/watchlisted/completed) are excluded from suggestions.

Context

Raw capture (Todoist, Druthers Backlog inbox): "Taste profile recommends titles and one click watchlist"

No existing recommendation engine was found in the druthers-api or druthers-web graphs — this is net-new. Existing building blocks that a recommendation feature would likely draw on: per-user rank/tracker data (DbUserMovie, DbUserTVShow, DbUserBook, DbUserVideoGame in druthers-api), and the existing watchlist add flow (movies/watchlist/page.tsx in druthers-web).

Estimate

• Recommended model: Opus 4.8 — new recommendation logic (what "taste profile" means, how to weight signals, how to avoid recommending already-tracked items) is architecturally open-ended, not a routine CRUD feature.
• Human effort: L — multiple product decisions needed (what counts as a taste signal, which domains get recommendations first, how many/how often to surface) plus review of the resulting recommendation quality.

Notes for Implementation (Optional)

Consider scoping the first version to a single domain (e.g. movies) before generalizing to TV/books/games, given the open questions below.