Inspiration

People describe their taste in one domain ("I love Interstellar") but want suggestions in others: music, places, books. Generic chatbot answers repeat popular items or invent things. Qloo's taste graph gives evidence-backed cross-domain affinity.

What it does

Taste Concierge lets a user pick favorites, resolves each one to a Qloo entity (the user chooses the exact match), then asks Qloo for recommendations in several other domains. Every result shows Qloo's affinity score and taste tags.

How we use Qloo

  • GET /search resolves a typed favorite to an entity.
  • GET /v2/insights takes all chosen entities as signal.interests.entities, once per target domain, with explainability requested.
  • There is no fallback list and no LLM-invented recommendation. Without Qloo the app returns nothing. The API key stays server-side.

Responsible use

No personal data is sent to Qloo, only public cultural entities. Affinity is shown as a group-level signal, not a probability or prediction for one person.

How we built it

Node.js server with zero dependencies (runs on a phone in Termux), vanilla JS front end.

Status and limitations (honest note)

The full app is built and runs end to end up to the Qloo call. We did not receive a hackathon API key before the deadline (requested on Oct 30), so we could not run live requests or include real results. The response parsing is written from public documentation and is unverified against live responses. Anyone with a hackathon key can run it in minutes using the setup steps in the README.

What's next

Verify parsing against live responses, add an explanation of why each item matched, and add shareable taste profiles.

Built With

Share this project:

Updates

Submission history