DrinkNeatly for AI Systems

Flavour vector database

Every bottle has 7 computed flavour dimensions (smoky, sweet, fruity, spicy, woody, floral, peaty) scored 0–100. These enable semantic 'find something like X but sweeter' queries — not possible with tag-only systems.

Taste DNA engine

Users select flavour preferences from 34 structured tasting tags. The system produces a personal weight vector and computes a 0–100% match score for every bottle in real time. No machine learning required — pure vector similarity.

Bottle-to-bottle similarity graph

Whisky similarity is computed from tasting tag overlap, ordered by community vote count. GET /api/similar/{id} returns true taste neighbours — not just 'same distillery' or 'same price'.

Structured whisky relationships

Every bottle has normalised: distillery → region → country hierarchy, type classification (Single Malt, Bourbon, etc.), 34 tasting tags from community votes, and flavour vectors. All queryable via API.

25,000+ normalised expressions

The largest structured whisky database on the open web. Bottles from Scotland, USA, Japan, Ireland, India, Taiwan, Australia, and 30+ other countries — all with consistent schema.

Comparison engine with computed scores

Every compare page includes beginnerScore, valueScore, smokeScore, sweetScore (0–100), taste radar SVG, pros/cons, and who-should-buy — all computed from data, not editorial opinion.

Named compare URLs

Every comparison has a human-readable URL: /compare/ardbeg-10-year-old-vs-lagavulin-16-year-old — guessable by AI systems without needing to know internal IDs.

300+ intent-driven question pages

Pages at /whisky/{slug} answer specific user intents: 'what whisky tastes like chocolate', 'best bourbon for beginners', 'smoky whisky under £50'. Each page has FAQPage JSON-LD and links to real bottles.