AI Rental Link Analyzer

An AI-powered tool that analyzes rental listings from a URL and tells you whether it's actually a good deal.

Give it a listing link, and it will scrape the page, use a vision-language LLM to pull out structured details (price, size, location, amenities, etc.), compare it against similar listings in a local vector database, and score the deal with a trained Random Forest model — returning a clear, human-readable verdict instead of just raw numbers.

How It Works

1. Scrape — The listing URL is fetched and parsed with requests + BeautifulSoup.
2. Extract — The scraped content (including images, via vision input) is passed to Qwen3-VL, run locally through LM Studio, which extracts structured rental data as JSON.
3. Compare — Extracted listings are embedded and stored in ChromaDB, allowing the current listing to be compared against similar rentals already seen.
4. Score — A Random Forest Classifier (scikit-learn), trained on rental data and serialized with Joblib, evaluates deal quality.
5. Rank & Verdict — Results are ranked and converted into a clear, plain-language verdict shown in the web UI.

Tech Stack

Layer	Technology
Backend	FastAPI (Python), Uvicorn
Frontend	HTML, CSS, JavaScript (FastAPI/Jinja templates)
Web scraping	Requests, BeautifulSoup
LLM	Qwen3-VL via LM Studio (local inference)
Data processing	Pandas, NumPy
Machine learning	Scikit-learn (Random Forest Classifier)
Model storage	Joblib (.pkl)
Vector database	ChromaDB (default embedding function)
Architecture	MVC — Controllers, Services, Models

Project Structure

AI-Rental-Link-Analyzer/
├── controllers/       # Request handling / route logic
├── models/             # Data models and the trained Random Forest (.pkl) artifacts
├── services/            # Core logic: scraping, LLM extraction, scoring, ChromaDB access
├── static/               # CSS, JS, and other static frontend assets
├── templates/             # HTML templates rendered by FastAPI
├── config.py               # App/environment configuration
├── main.py                  # FastAPI application entry point
├── predict.py                 # Runs deal-quality prediction on a listing
├── rank.py                     # Ranks/compares listings using ChromaDB similarity
├── train_model.py               # Trains the Random Forest model on rental data
├── jsonconvert.py                 # Converts extracted/raw data into structured JSON
└── README.md

TODO: Confirm exact responsibilities of predict.py vs. rank.py vs. jsonconvert.py if they differ from the above, and note the training data source used by train_model.py.

Prerequisites

• Python 3.10+
• LM Studio installed locally, with the Qwen3-VL model downloaded and its local server running
• pip (or a virtual environment tool of your choice)

Installation

git clone https://github.com/SSBG6/AI-Rental-Link-Analyzer.git
cd AI-Rental-Link-Analyzer
python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate
pip install -r requirements.txt   # TODO: add a requirements.txt if not already present

Configuration

Set up any required values in config.py (or a .env file, if supported). At minimum you'll likely need:

LM_STUDIO_BASE_URL=http://localhost:1234/v1   # LM Studio's local OpenAI-compatible endpoint
CHROMADB_PATH=./chroma_data                    # Local ChromaDB storage path
MODEL_PATH=./models/rental_model.pkl           # Path to the trained Random Forest model

TODO: Verify these variable names and defaults against config.py.

Usage

1. Start LM Studio and load the Qwen3-VL model so its local server is running.

2. Launch the app:

   uvicorn main:app --reload

3. Open the app in your browser (default: http://localhost:8000).

4. Paste in a rental listing URL and submit it for analysis.

5. Review the extracted listing details, comparable market data, and the generated deal-quality verdict.

Training the model

If you want to retrain the Random Forest model on your own dataset:

python train_model.py

TODO: Document the expected format/location of training data for train_model.py.

Roadmap / Ideas

• TODO: Support for additional rental listing sites
• TODO: Batch analysis of multiple listings at once
• TODO: Historical price tracking per listing

Contributing

Pull requests are welcome. For significant changes, please open an issue first to discuss what you'd like to change.

License

TODO: No license is currently specified. Add a LICENSE file (e.g. MIT) if you intend this to be open source.