Cultura: The AI Taste & Travel Assistant

The only AI that gets your vibe.

Comment

Inspiration

I’ve always felt like AI chatbots (from ChatGPT to Gemini) are great at general knowledge, but they fall short when it comes to taste.
Ask them for music or travel recommendations, and the answers are usually generic or surface-level.

That’s where I saw the gap:
There’s no assistant that truly understands your personal vibe.

Until now with Cultura.
Powered by Qloo’s Taste AI and Gemini, Cultura is built to give users tailored recommendations not just by domain, but by mood, personality, and preference.

What it does

Cultura is an AI-powered taste assistant that does two main things:

1. Taste Chat

It helps you discover what you actually like in movies, music, places, art, and more.
You simply enter a natural prompt like:

-I’m going to Italy and I love Wes Anderson films where should I go?-

Cultura uses Gemini + Qloo to understand your vibe, extract relevant taste tags, and recommend:

  • Cities to visit
  • Local experiences
  • Music, films, or art that match your taste
  • Cultural overlaps (e.g., film aesthetics & real-world places)

2. Trip Planner

Cultura also plans entire trips for you based on your preferences.

  • Step 1: You fill a form with budget, location, and travel days
  • Step 2: Cultura generates a full itinerary including:
    • Visa & airline options
    • Hotels and stay suggestions
    • Day-by-day activity breakdown

Personalization via chat:
I want to spend Day 3 in Shibuya → Cultura replans accordingly

The Vibe

Instead of tab-hopping between Google, Spotify, travel blogs, and group chats,
Cultura gives you taste-based recommendations in one place made for you.

How we built it

Cultura is built using a modular full-stack architecture with Next.js and TypeScript, powered by Qloo’s Taste API and Gemini Flash 2.5.

Tech Stack

Frontend:

  • Next.js 14 (App Router)
  • TypeScript
  • TailwindCSS for modern UI styling
  • React-based components for chat & planner flow

Backend:

  • API Routes using Next.js

APIs:

  • Qloo Hackathon API
  • Gemini Flash 2.5

Core logic:

  • Gemini parses the user query and extracts entities, locations, moods, and taste tags
  • Qloo endpoints like /tag/infer, /entity/resolve, and /recommendation are used to generate intelligent recommendations
  • The trip planner uses form inputs to:
    • Estimate trip cost
    • Suggest airlines and hotels
    • Break the trip into a daily itinerary
    • Let users tweak plans on the fly through chat
      Example: "Make Day 2 more chill"

Challenges we ran into

One of the biggest challenges was getting used to the Qloo API structure especially how to:

  1. Resolve ambiguous user input into valid Qloo entities or tags
  2. Chain endpoints like /tag/infer, /entity/resolve, and /recommendation effectively

Accomplishments that we're proud of

  • Successfully integrated Qloo and Gemini to create a personalized, taste-aware recommendation engine
  • Built a fully functional trip planner that generates detailed, editable itineraries from just a few inputs
  • Enabled natural language queries like "I'm going to Italy and like Wes Anderson films" to return curated, context-rich results
  • Designed a chat interface that feels like you’re talking to a real, culture-savvy assistant
  • Built and deployed the app from scratch while learning new APIs and tools on the fly

What we learned

  1. How to use Qloo’s API for tag inference, entity resolution, and multi-domain recommendations
  2. Prompt engineering for Gemini to extract structured data from messy user inputs
  3. Managing stateful conversations in a dynamic chat interface

What's next for Cultura: The AI Taste & Travel Assistant

  • Multi-modal support (voice input, image-based moodboards, Spotify/Letterboxd profile parsing)
  • Deep integration with Spotify, Google Maps, YouTube, and booking platforms

GitHub

Built With

Share this project:

Updates