Inspiration
Every group chat has that one endless debate: "kya karein?" Movie? Dinner? Music? Someone wants action, someone wants biryani, someone wants lo-fi. We built SquadVibe to end that fight forever — one AI-powered plan everyone actually vibes with.
What it does
SquadVibe lets you create a squad plan, share a code with friends, and have everyone add 2–3 favorites (movies, artists, cuisines). Our Taste Blend engine queries Qloo's taste graph (250M+ entities) per member, then blends signals — genres, keywords, popularity, entity coverage — into one perfect plan: the movie, the restaurant, and the music your whole squad will love. Includes consensus/gems/wildcard variants, a 3-stop itinerary, Taste DNA visualization, and squad voting.
How we built it
- Backend: Express + TypeScript, deployed on Render (Singapore)
- Frontend: Vite + React + Tailwind
- Taste Blend engine: Qloo Search API with per-member round-robin queries, typed entity resolution, core-genre gating, member-coverage scoring, location-biased restaurant search, and request throttling (4 concurrent, 150ms spacing)
- We replaced Qloo's broken Insights API mid-build with a Search-powered blend engine — honest engineering over fake claims
Challenges we ran into
Qloo's Insights API returned errors, so we rebuilt the entire recommendation engine on the Search API in one sprint. We also fought cross-category contamination (movie signals generating restaurant picks) and fixed it with strict category gating. Location consistency was another battle — foreign restaurant entities sneaking into local results.
What's next
Locked demo datasets for judges, real-time squad collaboration, and Spotify/YouTube playlist export. dikhao!
Built With
- express.js
- qloo
- react
- render
- rest-api
- tailwind-css
- typescript
- vite