Inspiration
We’ve both landed in cities we didn’t know and opened the same apps everyone else uses. The spots are fine. They’re just not ours. You might care about hand-painted shop fronts and quiet back streets because of the films you watch, or about a café’s playlist because of what you listen to on repeat. Trip lists don’t see any of that. They optimize for popularity, not taste.
The hackathon pushed us toward Qloo, and that clicked. Your music, movies, and moods aren’t separate hobbies. They’re one profile. What we wanted was the thing a friend who knows you would text before a trip. Not “here are ten famous places.” More like “you’ll like this kissaten. It feels like the slow scenes in the movies you love.”
That’s VibeGuide. You pick a city, name a few things you actually care about, and get a small day plan. Each stop has a short line about why it fits you. Recommendations are useful. Feeling understood in a foreign city is the part we were chasing.
How we built it
FastAPI on the backend, React and Tailwind on the front. You pick a city on a map, type a few anchors, hit generate. Qloo search resolves those, insights finds venues in that city, and we fill morning through night with four stops. Gemini writes the blurbs from Qloo’s data. If there’s no key, templates kick in. Tokyo and NYC have offline fallbacks so a demo doesn’t die mid-pitch.
Armita did the UI and product flow. Chinmoy wired Qloo and the itinerary pipeline.
Challenges
Entity resolution mattered more than we expected. Keywords alone gave weak insights. Latency is real (a few API calls per run), so we cache Qloo responses and show simple loading messages. This was our first full-stack project together, so we stuck to one screen and one path that has to work live.
What's next
Shareable itineraries, better refine after thumbs up/down, and more cities only when the Qloo data is solid enough to trust.
