Mood-aware movie and anime recommender with "why this fits" explanations

The agent builds an always-on picture of the user's current mood. On every request, it pulls recent context from the last few chat messages and any journal entries, then combines that with whatever the user just said (free text, emoji, or both) to infer the mood. It then searches a catalog of popular movies and anime (pre-seeded metadata: plots, tags, genres, mood signals) and, if the catalog feels thin, augments the result set with a fresh web pull via the Web search or Brave Search API for currently-trending or region-relevant titles. The user's "where to watch" region is set during first-run setup. It re-ranks candidates with a local LLM for best mood fit, returns 3–5 picks (default mix of movies + anime, user can override per request), and for each title gives a warm one-line "why this matches your mood" reason plus the year, genres, a short teaser, the rating, and a "where to watch" hint. The agent remembers every pick the user likes, skips, or finishes, and uses that history to refine future suggestions. Optional extras like a pixel-art mood poster for the pick are available on request.