🏆 2nd place — Cursor Hackathon 2026
TableUs is a location-aware restaurant planning demo for small groups.
The project centers on a straightforward flow: choose a location, describe what the group wants in plain language, and turn a nearby restaurant pool into a short ranked list with concise reasoning. The app uses demo users, in-memory data, Google Maps, and Gemini to prototype that decision-making loop.
- Small-group meal planning rather than long-term social networking
- Nearby restaurant discovery grounded in real geocoding and Places data
- Natural-language search with explainable recommendations
- Demo-friendly preference summaries that make solo and multi-person search easier to test
- Fast local setup with no database required
The repository includes lightweight user connections, demo profiles, and review-derived preference text, but those pieces exist to support the planning flow in a no-login prototype. The main product question in this repo is simple: how can a group narrow choices faster when location, cuisine, price, and atmosphere all matter at once?
- Pick a demo user from the sidebar.
- Set a location by searching for a place name.
- Browse nearby restaurants pulled from Google Places.
- Enter a natural-language search such as
casual ramen near downtownorquiet dinner spot for three. - Optionally include additional demo users to run a multi-person search across several preference summaries.
- Review the ranked results and the short AI reasoning attached to each result.
You can also submit a natural-language review or upload a food photo to test the supporting profile and analysis flows.
- Location resolve via geocoding before search
- Nearby restaurant candidate pool from Google Places
- Solo restaurant ranking from query plus taste summary
- Group ranking from combined demo-user preferences
- Natural-language reviews that refresh stored preference text
- Food photo analysis through Gemini Vision
- Demo friends management for testing group scenarios
- In-memory demo data so the app runs without a database
TableUs is best understood as a planning tool prototype.
- It is location-first: recommendations are tied to a real place and a live nearby pool.
- It is decision-oriented: the goal is to narrow options, not build a permanent foodie identity layer.
- It is explainable: each top result comes back with short reasoning rather than a black-box score alone.
- It is intentionally lightweight: demo users and in-memory state keep the build easy to run and easy to iterate on.
| Layer | Choice |
|---|---|
| Frontend | Next.js 16, React 19, Tailwind CSS 4, Framer Motion |
| Backend | FastAPI, Python |
| AI | Google Gemini for vision, ranking, and preference merging |
| Maps | Google Maps Geocoding API and Places Nearby Search |
frontend/- Next.js app for discover, connections, review, and profile flowsbackend/- FastAPI API for maps lookup, search, reviews, food analysis, and demo data access
Main frontend routes:
/discover- location-aware restaurant search/friends- demo connections and combined planning state/review- natural-language review submission/profile- current demo user's preference summary