QlooSense AI — Taste-Aware City Discovery

Inspiration

Finding places in a city can be difficult when recommendations do not reflect a person's interests. QlooSense AI explores how taste-aware discovery can help people find relevant local experiences by combining an interest with a city.

What it does

QlooSense AI is a taste-aware city discovery demo powered by Qloo's API. Users enter an interest, such as art gallery, and a city, such as Surat. The application sends the request to a deployed backend and displays the venue information returned by the API.

The demo also provides a transparent comparison with a simple baseline and displays execution logs and request metrics so reviewers can inspect the workflow.

How we built it

Frontend: HTML, CSS, and JavaScript for the interactive comparison interface.
Backend: Node.js and Express.js, written in TypeScript, deployed on Vercel.
Discovery: Qloo API integration through a server-side backend endpoint.
Deployment: Vercel for the frontend and backend, with separate GitHub repositories.

The frontend communicates with the backend through a POST request. The Qloo API credential stays in the backend's server-side environment rather than being exposed in browser code.

Challenges we faced

One important challenge was connecting a static frontend to a deployed API while keeping credentials secure. Another was making the comparison understandable without presenting the baseline as a real AI-generated recommendation. We addressed these by separating the deployments, keeping the API key server-side, displaying execution logs, and clearly explaining the baseline's limitations.

What we learned

This project helped us explore API integration, deployment, request tracing, and transparent presentation of recommendation results. It also reinforced the importance of distinguishing returned API data from assumptions and avoiding invented venue information when results are unavailable.

What's next

Future improvements could include richer preference inputs, more evaluation examples, and a separately integrated language model for a clearly labeled LLM-versus-Qloo comparison.