Skip to content

Claude/pensive pascal pe ng r - #74

Merged
Athenox14 merged 2 commits into
masterfrom
claude/pensive-pascal-PENgR
Jun 3, 2026
Merged

Claude/pensive pascal pe ng r#74
Athenox14 merged 2 commits into
masterfrom
claude/pensive-pascal-PENgR

Conversation

@Athenox14

Copy link
Copy Markdown
Owner

No description provided.

claude added 2 commits June 3, 2026 22:48
Add an end-to-end watchlist feature and substantially broaden the
recommendation algorithm's inputs.

Watchlist:
- new watchlist table + toggle/status/getWatchlist tRPC endpoints
- detail-page toggle button, home 'My list' rail, i18n (fr/en/de)
- WATCHLIST_ADD/REMOVE analytics events; watchlist folded into the genre
  taste model and used as a positive affinity signal in scoring

New signals fed into the algorithm (scoreCandidateWithProfile):
- likes / dislikes (strengthened), watchlist affinity
- per-title abandonment curve (normalized deciles), effective completion
  (>75%), sessions-to-finish, impression/hover click-through — via a new
  cross-user media_stats table + mediaStatsEngine
- active vs idle playback time
- impressions-without-click and hover-without-open
- collection / saga, keywords / themes, director / composer, age rating,
  popularity / novelty (new TMDB-enriched media columns)
- genre x moment (weekday x daypart affinity)
- binge / session detection + co-viewing capture
- seasonality (month buckets) and churn (gaps, reactivations)

Data model:
- media: keywords, director, composer, certification, popularity
- new tables watchlist, media_stats (idempotent migrations)
- TMDB enrichment for keywords, crew, certification, popularity

Docs: docs/recommendation-algorithm.md documents every signal and lists
proposed future improvements (cold-start, recency decay, diversity/MMR,
score normalization, offline eval, etc.).

https://claude.ai/code/session_01RF3jY6P1ySdMQzuv97NFtL
Document the signals added in this PR that are now observed and fed into the
recommendation algorithm: explicit watchlist, per-title playback quality
(abandonment curve, effective completion >75%, sessions-to-finish, impression/
hover CTR), active vs idle time, content metadata (collection, keywords,
director, composer, certification, popularity, novelty), genre x moment,
seasonality, churn, and binge/co-viewing.

https://claude.ai/code/session_01RF3jY6P1ySdMQzuv97NFtL
@Athenox14
Athenox14 merged commit 76d82cc into master Jun 3, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants