Predict the next pour.

The fifth pillar of x-brain — cannabis (BudBrain), wine (WineBrain), food (FoodBrain), beer (BeerBrain), and now coffee. Same engine, coffee-domain features.

What CoffeeBrain does

• Snap a bag, get a prediction. The roaster bag's front label extracts producer, coffee name, origin, processing, roast date, and infers perceptual feature scores (acidity, body, sweetness, bitterness, tasting notes).
• Snap a café menu, get N predictions. The chalkboard at your local third-wave café becomes a ranked list of pourovers + espresso drinks scored against your taste fingerprint.
• Origin-aware similarity. Cosine over roast/acidity/body plus indicator features (origin_ethiopia, processing_natural, varietal_geisha) keeps a Sumatran natural from being recommended after you only loved Ethiopian washed lots.
• Roaster context matters. Same Ethiopian Yirgacheffe, two different roasters — the predictor learns your preference per producer.
• Cross-brain pairings. Friends who loved this Geisha pourover ate this almond croissant; espresso + biscotti is suggested in the morning if your cohort eats biscotti at brunch. Powered by the x-brain knowledge graph.

How the prediction works

Same hybrid engine that powers BudBrain, WineBrain, FoodBrain, and BeerBrain: a 15-question style-discriminator onboarding ladder, k-means cohort assignment, content-based fallback over the coffee feature schema, and a per-user XGBoost model that takes over after ~20 ratings.

Where the catalog comes from

The Philadelphia / SE PA specialty-roaster lineup is seeded — La Colombe, Elixr, Ultimo, ReAnimator, Saxbys, Function Coffee Labs, Ox, Square One, Menagerie, Vault + Vine, Spruce Street Espresso, Greenstreet, plus regional anchors in West Chester (Higher Grounds) and Phoenixville (Volo). Single-origin anchors cover Ethiopia / Colombia / Kenya / Guatemala / Brazil / Sumatra / Panama Geisha so cohort assignment has dynamic range. Sweet Maria's open cupping logs and Coffee Review professional scores extend the catalog nightly. Per-roaster website scrapers fill in detail rows for the SE PA flagships. Anything missing? Snap it — the bag-vision flow seeds new items as you encounter them.