CommonGround — Different Tastes, One Great Plan

Inspiration

Great experiences are rarely a one-person decision. Most recommendation engines optimize for a single user; CommonGround instead finds experiences that make sense for two people with different cultural tastes.

What it does

Two people enter cultural favorites—films, music, artists, books, or brands—and choose what they want to discover together. The agent is designed to resolve these favorites against Qloo cultural entities, use Qloo Insights to discover meaningful matches across categories, and rerank suggestions to avoid one person's preferences dominating. It presents a balance index, evidence-based explanations, and a visible reasoning trail.

How we built it

Responsive HTML/CSS/JavaScript UI, serverless Node.js agent orchestrator, and integrations for Qloo /search and /v2/insights. The planner separates entity resolution, parallel per-person and joint affinity retrieval, fairness reranking, and explanation generation. Credentials stay server-side.

Qloo's role

Qloo is indispensable: real candidates and the cultural associations underpinning them are sourced from Qloo's taste graph. Our ranking layer adds fairness across two profiles; it does not invent Qloo cultural matches. When credentials are missing, the app clearly labels synthetic illustrative preview output rather than pretending it is live Qloo data.

Challenges and lessons

Building a two-person recommender requires balancing relevance and fairness without conflating an app-computed score with Qloo's native affinity signal. We made provenance explicit and kept preview mode transparent.

Current status

Publicly deployed application and open-source code are available. Live Qloo requests require QLOO_API_KEY configured on Vercel; until then, only the illustrative preview operates. The project should not be treated as a fully verified live-Qloo submission until API-backed requests are tested.

Next steps

Validate Qloo authentication and endpoint compatibility with a real key, verify cross-person recommendation results, and perform accessibility and edge-case QA.

Author: Samer Tallauze.

Built With

• agentic-ai
• css3
• html5
• javascript
• node.js
• qloo
• serverless
• vercel