SceneSet

Search the culture. Find the city. Prove the recommendation

Comment

SceneSet

Search the culture. Find the city. Prove the recommendation.

Inspiration

Finding a city's cultural scene shouldn't depend on generic travel recommendations or an AI's best guess.

Whether you're searching for Afrobeats, late-night jazz, or the perfect food-and-music atmosphere, traditional discovery platforms often prioritize popularity over cultural compatibility.

SceneSet was created to answer a different question:

Which city actually matches the culture I'm looking for, and what evidence supports that recommendation?

What It Does

SceneSet is a Qloo-powered cultural discovery agent that helps people find taste-aligned places across different cities.

Instead of generating unsupported suggestions, SceneSet uses real cultural intelligence to identify relevant places, compare cities, and explain its decisions.

Users can:

  • Describe their desired cultural scene in natural language.
  • Compare two to five cities.
  • Discover places using Qloo's cultural taste intelligence.
  • Explore ranked city comparisons and venue recommendations.
  • Inspect the taste signals behind each recommendation.
  • Generate a verifiable SHA-256 Taste Receipt.

The result is a transparent, evidence-backed approach to cultural discovery.

How We Built It

SceneSet combines Qloo's Taste AI API with an agent workflow powered by Eve.

The process begins when a user submits a cultural query and selects candidate cities.

First, the agent uses Qloo's Tags API to resolve the query into meaningful cultural taste signals.

Next, those signals are supplied to Qloo's Insights API to retrieve relevant places and affinity values in each selected city.

SceneSet then applies a consistent, versioned scoring algorithm to compare the markets.

Finally, the system produces a ranked recommendation and a SHA-256 Taste Receipt containing the evidence behind its decision.

The technical architecture includes:

  • React and Vite for the frontend.
  • TypeScript and Nitro for backend services.
  • Eve for agent tooling and execution.
  • Qloo Tags and Insights APIs for cultural intelligence.
  • Zod for validation.
  • Vitest for automated testing.
  • Vercel for production deployment.

Why Qloo Matters

Qloo is central to SceneSet's functionality.

Its taste graph provides the cultural signals, place entities, and affinity values used to make recommendations.

Without Qloo, the agent would lack the external cultural evidence required to produce its comparisons.

SceneSet deliberately avoids inventing recommendations when provider information is unavailable.

This creates an important distinction between AI-generated suggestions and culturally grounded decisions.

Challenges We Faced

One challenge was converting open-ended cultural interests into structured taste signals.

Another was ensuring that different cities could be compared fairly using a consistent scoring approach.

We also wanted users to understand why a particular city was selected.

To address these challenges, we implemented deterministic ranking, explicit limitations, provider-request evidence, and verifiable recommendation receipts.

Accomplishments

SceneSet progressed from a cultural discovery concept to a publicly deployed application with a working Qloo integration.

Our documented verification included:

  • Eight passing deterministic tests.
  • A successful credentialed Qloo integration test.
  • Successful frontend and backend production builds.
  • A working three-city comparison.
  • Four Qloo provider calls in the documented example.
  • Verified receipt integrity.
  • A successful anonymous production browser walkthrough.

What We Learned

Cultural recommendations become more useful when users can inspect the evidence behind them.

We also learned the importance of distinguishing cultural affinity from event availability.

SceneSet recommends taste-aligned places, but it does not claim to verify scheduled events, ticket availability, opening hours, or reservations.

What's Next

We want to expand SceneSet with richer personalization, more cultural discovery categories, and additional ways for users to explore the reasoning behind recommendations.

Our long-term vision is to make culturally intelligent discovery more transparent, personalized, and useful across cities worldwide.

SceneSet — Search the culture. Find the city. Prove the recommendation.

Built With

Share this project:

Updates

Submission history