Taste Twin

• 

  Taste Twin landing screen

• 

  A sample weekend plan generated in demo mode (synthetic data)

• 

  The Taste Receipt from a live Lisbon test, resolving favourites to Qloo entity IDs

Inspiration

Weekend recommendations are usually generic top-ten lists that ignore who is travelling. Taste Twin starts from a person's actual favourites and treats every suggestion as a claim that should show its evidence.

What it does

Tell Taste Twin a handful of your favourites - a film, an artist, a book, a dish - and pick a city. It builds a Friday-to-Sunday plan: a cocktail bar, a restaurant, an artist to hear, a book for the trip and a cafe for the morning after. Every pick comes with a receipt showing which favourite it follows, the search steps taken and the taste signals behind it, so the reasoning stays inspectable. Day intros are narrated by an open model.

How we built it

Next.js 16, React 19, TypeScript and Tailwind, deployed on Vercel. A taste adapter resolves free-text favourites through Qloo's search API, then queries Qloo's recommendation API across domains - film, music, books and food - and builds each receipt from the entity IDs and signals in the response. Day intros are narrated by an open model (Nemotron) hosted on Nebius. With no API keys configured, the app runs on a clearly labelled synthetic sample.

Challenges we ran into

Resolving free-text favourites to the right taste entities across domains, and keeping every recommendation's reasoning inspectable instead of buried in a single opaque answer.

Accomplishments that we're proud of

The full loop works end to end: four favourites became a complete Lisbon weekend of real places in testing, and every pick carries its evidence. With no keys configured, the app still demonstrates the whole flow on a labelled sample.

What we learned

Cross-domain taste affinities - film to music to food - can anchor a coherent itinerary when the query chain stays visible. Showing evidence turns a list of names into something a traveller can check.

What's next for Taste Twin

Richer plans across more of a trip, deeper evidence for each pick, and taste profiles that carry across weekends.

Built With

• nebius
• next.js
• qloo
• react
• tailwind
• typescript
• vercel