rented.

Renting in Toronto shouldn't be this hard.

Comment

Renting in Toronto shouldn't be this hard.

Python React Flask Vite BeautifulSoup

The story behind this

I'm the son of a landlord.

My dad doesn't really like that word — "landlord" — and honestly, neither do I. He's just someone who bought a house, couldn't afford the mortgage alone, and had to rent part of it out. Every month, I'm the one writing the Kijiji listings, fielding the calls, and trying to find a decent tenant before the mortgage is due. I've sat on both sides of this thing.

And from where I sit, the rental market is genuinely broken — not just for tenants, but for everyone in it.

Renters are scrolling through pages of overpriced listings, duplicate ads from the same landlord, promoted posts they didn't ask for, and buried in half of them — illegal clauses. Things like "no pets", "students only", "no subletting" — restrictions that are straight-up violations of the Ontario Residential Tenancies Act. Most people never catch them.

But this isn't a tale about them. It's about the potential tenants — the next generation, my generation — who may not have a home in the country they call home.

Meet rented.

What it does

Enter what you're looking for — city, price range, bedrooms, bathrooms. rented. scrapes Kijiji live, scores every result against your filters, flags anything with illegal clauses, and returns a clean carousel of your top 10 matches.

No ads. No promoted listings. No noise.

Just the places that actually fit.

Features

  • Live Kijiji scraping — hits 3 pages of GTA listings in real time
  • RTA violation detection — scans descriptions for illegal clauses across 7 categories
  • Smart scoring — ranks listings by how well they match your filters
  • Animated loader — a hand-crafted SVG house builds itself while you wait
  • Centred carousel — your top match front and centre, neighbours fading on both sides
  • Ad-free, always — no promoted posts, no noise

How it works

1. Scraping

kijiji_V2.py hits the first 3 pages of Kijiji's GTA apartments & condos section. For each listing card it pulls title, price, location, description, the lead photo, and structured attributes — bedrooms, bathrooms, unit type, parking, and square footage — directly from Kijiji's own attribute list.

Price is cleaned from a raw string like $1,750/mo down to a plain number. Anything that isn't a dollar amount (e.g. "Please Contact") becomes 0 and gets filtered out downstream.

2. RTA violation detection

This is the part I care about most. Before a listing ever reaches you, its description is scanned against a library of regex patterns organised by violation type:

Category Examples caught
Gender restriction female only, looking for male, must be female
Ethnicity / religion indian only, muslim only, prefer asian
Age / occupation students only, no students, working professionals only
Pets no pets, no dogs, pet free home
Guests no guests, no overnight guests, visitors not allowed
Subletting no subletting, no sublease
Family status no couples, no families, single person only

Every pattern that matches gets added to the listing's illegal_flags list. Flagged listings aren't removed — they show up with a visible warning badge so you can make an informed decision. Transparency over gatekeeping.

Current limitation: the scraper only checks the short preview description on the listing card, not the full listing page. A deeper per-listing scrape would catch far more. That's next.

3. Scoring

Once listings are filtered by location and price range, each one is scored:

  • Exact bedroom match → +20 pts
  • Bathroom match or better → +20 pts
  • Has a real photo → +20 pts
  • Price at your minimum → +40 pts
  • Illegal RTA clause found → −20 pts per flag

Top 10 by score are returned. The rest are dropped.

4. The UI

Results come back through a Flask REST API and land in a React carousel. Your top match sits front and centre at full size. Cards on either side scale down and blur slightly, giving depth. Clicking any side card brings it forward. A dot nav and arrow buttons let you move through the full list. While results load, a small SVG house animates itself being built — walls, roof, chimney, smoke and all.

Project structure

rented/
├── api.py              # Flask server — exposes /search
├── kijiji_V2.py        # Kijiji scraper → listings.json
├── listings.json       # Cached listings
├── requirements.txt
│
└── rented-ui/          # React + Vite frontend
    └── src/
        ├── App.jsx     # Everything — home, loader, carousel
        └── main.jsx

Running it locally

You'll need Python 3.8+ and Node.js 18+.

Backend

pip install flask flask-cors requests beautifulsoup4
python api.py
# → http://127.0.0.1:5000

Frontend (in a second terminal)

cd rented-ui
npm install --legacy-peer-deps
npm run dev
# → http://localhost:5173

Both need to be running at the same time. Flask handles the data, Vite serves the UI.

Deploying

Part Platform Notes
Backend Render Add gunicorn to requirements.txt, start command: gunicorn api:app
Frontend Vercel Auto-detects Vite, one-click deploy

Update the fetch URL in App.jsx to your Render backend URL before deploying the frontend. Render's free tier spins down after 15 minutes of inactivity — first request after idle may take ~30s to wake up.

API

POST /search

{
  "location":  "Toronto",
  "minPrice":  800,
  "maxPrice":  2500,
  "bedrooms":  "2",
  "bathrooms": "1"
}

Returns an array of up to 10 listings sorted by score. Each listing includes title, price, location, bedrooms, bathrooms, sqft, parking, illegal flags, image, URL, and score.

What's still broken / what's next

  • [ ] date_posted always returns null — the scraper isn't catching it yet
  • [ ] RTA detection only checks the short preview text — a per-listing deep scrape would be far more accurate
  • [ ] Only Kijiji for now — Facebook Marketplace and Realtor.ca are next
  • [ ] Scoring weights are hardcoded — ideally user-adjustable
  • [ ] No saved searches, accounts, or history yet

Built at Hack Canada  ·  Made with 🍁 and an unreasonable amount of BeautifulSoup(requests.get(kijiji))

Share this project:

Updates