🏡 AI Property Hunter

By Shawn Strasser

🔗 Live Demo: house.up.railway.app

---

A smart property analysis tool that uses AI to evaluate real estate listings the way I would - checking everything from kitchen quality to crime rates. It scrapes Zillow, analyzes photos with Google's Gemini AI, and scores properties based on 25+ factors I care about. Plus, there's a mobile-friendly website where my wife and I can browse through properties and save our favorites.

---

🎯 The Cool Parts

AI that sees what you see

• Google Gemini 2.5 Pro turns photos + text into 25+ human-centered features
• Clear ratings for kitchen, baths, privacy, views, road exposure, and more

Scoring that blends real-world data

• FBI crime stats, drive time to grocery, and school ratings
• Smart normalization + customizable weights → one meaningful total score

Built to actually use

• Mobile web app to browse, rate (Yes/Maybe/No), and add notes
• Production-ready, cloud-backed ratings; simple, private, and fast

---

🛠️ Data Pipeline

┌─────────┐    ┌──────────────────┐    ┌─────────────────┐
│ Zillow  │───▶│ Playwright       │───▶│ Gemini AI       │
│ Listings│    │ Scraper          │    │ Analysis        │
└─────────┘    └──────────────────┘    └─────────────────┘
                                                │
                                                ▼
┌─────────────┐                        ┌──────────────┐
│ Crime Data  │─────────────── ───────▶│ DuckDB       │
└─────────────┘                        │ Scoring      │
                                       │              │
┌─────────────┐    ┌──────────────┐───▶│              │
│ AI Results  │───▶│ Maps API     │    │              │
│ (Addresses) │    │ (Grocery     │    │              │
└─────────────┘    │  Stores)     │    └──────────────┘
                   └──────────────┘             │
                                                ▼
                                       ┌─────────────────┐
                                       │ Flask Web App   │
                                       └─────────────────┘
                                                │
                                                ▼
                                       ┌─────────────────┐
                                       │ Railway Deploy  │
                                       └─────────────────┘

---

---

📁 Project Structure

House/
├── flask_app.py           # Flask app (UI/API, SQLite Cloud ratings)
├── live_scraper.py        # Zillow scraper + Gemini feature extraction
├── feature_extraction.py  # Gemini prompts/tools + features table schema
├── scoring.py             # Scoring SQL and helpers
├── nearest_grocery.py     # Google Maps: nearest grocery + drive time → grocery table
├── config.py              # Weights, scoring params, app config, emojis
├── templates/
│   ├── base.html
│   ├── index.html
│   ├── settings.html
│   └── error.html
├── crime.ipynb            # Populate crime data (creates crime table)
├── requirements.txt
├── railway.toml
└── property_data.db       # DuckDB file (created at runtime by scripts)

---

🚀 How to Use

1. Update the Zillow search URL

• Edit live_scraper.py and change search_url to your target area/filters.

2. Set environment variables

• GEMINI_API_KEY (Google Generative AI)
• MAPS_API_KEY (Google Maps Platform)
• SECRET_KEY and APP_PASSWORD (Flask app)
• RATINGS_DB_URL (SQLite Cloud URL for ratings/notes)

3. Run the scraper

• python live_scraper.py (creates/updates property_data.db with listings + AI features)

4. Add crime data

• Open crime.ipynb and run it to populate the crime table.

5. Add grocery stores

• python nearest_grocery.py (uses MAPS_API_KEY to fill the grocery table with nearest store + drive time)

6. Deploy to Railway

• Push the repo, connect on Railway, set the env vars above, and deploy