Find your place — not just an apartment.

Canopi is an AI-powered rental discovery platform that understands who you are, not just what you're searching for. Through lifestyle-revealing conversation, it learns your priorities and surfaces rentals across Canada that actually fit your life.

Live demo: https://hack-canada.vercel.app

Every rental platform in Canada filters by beds and price. They treat you like a set of constraints. Canopi starts with you — your habits, your energy, what a good day actually looks like — and works backwards to find where you belong.

An AI assistant asks indirect, personality-driven questions — "What does a good Sunday morning look like for you?" — and infers your preferences across 8 lifestyle axes: walkability, nourishment, wellness, greenery, buzz, essentials, safety, and transit. As the conversation evolves, the map re-ranks listings in real time to match your actual life.

• AI chat assistant — Conversational matching that reads between the lines. Supports English and French. Voice input/output via ElevenLabs STT/TTS.
• Interactive map — Mapbox GL map showing 200+ real Canadian rentals with price pins, listing cards, and fly-to animations when the AI recommends a property.
• 8-axis preference radar — Live spider chart that updates as Canopi learns what matters to you.
• Neighborhood scores — Amenity counts (schools, cafés, parks, groceries, transit, pharmacies, restaurants) within 1 km of every listing.
• 3D diorama view — Three.js spatial visualization of neighborhood vitality around a selected listing.
• Saved listings — Bookmark favorites with Supabase-backed persistence across sessions.
• Auth — Email/password and Google OAuth via Supabase.

Getting structured, enriched rental data at scale was one of our core technical challenges. We used a multi-stage AI-assisted pipeline:

1. Scraping — We pulled raw rental listings from RentFaster.ca, collecting addresses, prices, unit details, and geographic coordinates across Canadian cities.

2. AI parsing & normalization — Raw listing data is inconsistent: mixed formats, missing fields, non-standard descriptions. We used Gemini to parse and normalize listing text into a consistent schema — extracting bedroom counts, amenity mentions, building types, and income thresholds even when listings didn't follow any standard format.

3. Geospatial enrichment — For each listing, we queried the Overpass API (OpenStreetMap) to count nearby amenities within a 1 km radius: cafés, gyms, parks, grocery stores, pharmacies, schools, transit stops, and restaurants. This turned raw coordinates into neighborhood lifestyle scores.

4. Caching — Enriched data is cached locally (data/geoapify-places-cache.json) to avoid redundant API calls and keep the app fast.

The result: 200+ listings each carrying a rich neighborhood profile that the AI can reason about when making personalized recommendations.

git clone <repo-url>
cd hackcanada
npm install

Create .env.local:

NEXT_PUBLIC_MAPBOX_TOKEN=your_mapbox_token
GEMINI_API_KEY=your_gemini_key
NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
ELEVENLABS_API_KEY=your_elevenlabs_key

npm run dev

Open http://localhost:3000.

src/
├── app/
│   ├── page.tsx                  # Main map + chat interface
│   ├── saved/page.tsx            # Saved listings gallery
│   ├── diorama/page.tsx          # 3D neighborhood view
│   └── api/
│       ├── chat/route.ts         # Gemini conversational AI
│       ├── listings/route.ts     # Rental listing data
│       ├── vitality/route.ts     # Amenity data via Overpass
│       ├── tts/route.ts          # ElevenLabs text-to-speech
│       └── stt/route.ts          # ElevenLabs speech-to-text
├── components/
│   ├── avenuex/                  # UI components (map, chat, navbar, spider chart)
│   └── three/                    # 3D diorama components
└── lib/
    ├── spider-prefs-context.tsx  # 8-axis preference state
    ├── auth-context.tsx          # Supabase auth provider
    └── avenuex-data.ts           # Listing types and scoring
data/
├── rentfaster-listings.combined.json      # ~200 Canadian rental listings
├── rentfaster-listings.livable-data.json  # AI-enriched listings for chat
└── geoapify-places-cache.json             # Cached OSM amenity data
scripts/
├── enrich-rentfaster-listings-with-places.mjs   # Geospatial enrichment
└── clean-combined-listings.mjs                  # AI normalization pass

npm run enrich:places   # Pull amenity data from Overpass API
npm run clean:combined  # Normalize and AI-parse raw listings