Film BrainAI Film-Library Brain

500+ fixed genre nodes reshaped into a 14-dimension flexible taxonomy — editors find films in one sentence; the AI is honest when nothing matches and explains itself when it does.

How it works

Search replay

GitHub

Why now: Netflix is betting on generative-AI "find a film in one sentence" — with examples like "a crime drama like Money Heist but not too bloody" and "an easy comedy for a relaxing weekend", almost exactly the queries here. The difference: they're testing it; this hackathon prototype already built it. — Shang Media, 2026-06-05

🔍

Semantic search

More than keywords — describe what you want in one sentence and it finds films by intent.

🏷️

Auto-tagging

AI picks tags from a 14-dim / 400-tag taxonomy; whitelist-validated, editor-reviewed.

✅

Honest match

When nothing truly matches, it says so — no more relative scores faking a perfect 100%.

💬

Explainable results

Every film states why it surfaced — which conditions it matched, and how it was found.

🏆

Award tracking

Nomination lists from 18 award sites — auto-scraped, AI-structured, Wikidata-verified, matched to library tags.

📈

Self-eval loop

An LLM-judge harness over 45 real queries drives tuning — winners stay, losers roll back, nDCG@5 0.93 → 0.96.