📦 CREA MLS RAG Playground

A minimal yet practical RAG system built around CREA MLS-style listings, using:

• Python

• FAISS (vector search)

• FastAPI

• OpenAI embeddings + chat models

This repo is built as a companion to the article:

👉 RAG Explained: Engineering Scalable Search Without Hallucinations
https://nidly.substack.com/p/rag-explained-engineering-scalable?r=a3p8i

If you're coming from that post, this is the hands-on demo.

This project gives you a simple but realistic end-to-end pipeline:

1. Fake CREA listings (structured JSON)

2. Text → Embedding conversion

3. FAISS index for similarity search

4. Retrieval pipeline (Top-K)

5. RAG assembly layer for clean prompts

6. FastAPI endpoint: /ask

All in less than 300 lines of code.
No frameworks. No junk. Just engineering.

crea-mls-rag-playground/
  ├── data/
  │    └── listings.json        ← sample CREA-style listings
  ├── src/
  │    ├── config.py            ← env, config
  │    ├── embeddings.py        ← embedding helper
  │    ├── vector_store.py      ← FAISS index builder & search
  │    ├── ingest_crea_listings.py
  │    ├── rag.py               ← retrieval + generation
  │    └── api.py               ← FastAPI server
  ├── requirements.txt
  ├── README.md   (this file)

Install deps:

pip install -r requirements.txt

Create embeddings + FAISS index:

python -m src.ingest_crea_listings

Run API:

uvicorn src.api:app --reload

Query:

curl -X POST http://localhost:8000/ask \
  -H "Content-Type: application/json" \
  -d '{"query": "show me condos with ocean view"}'

Real estate has:

• Rich structured metadata

• Semi-structured descriptions

• High-variance natural language

• Real-world need for relevance ranking

It’s the perfect playground for understanding embeddings, vector-search, and RAG…
and honestly،

Built by Alireza Rahmani Khalili
Principal Software Engineer • AI Engineering • DDD • MLS Systems

• Substack: https://nidly.substack.com

• LinkedIn: https://www.linkedin.com/in/rahmanikhalili/

• Personal Website: https://alireza.engineer

Alireza writes about:

• AI engineering

• Vector databases

• Domain-Driven Design

• Real-estate listing systems

• Building nation-scale data pipelines

If you're into engineering that actually works in production, subscribe.