Real Estate RAG API

A multi-tenant Retrieval Augmented Generation API for real estate teams. Each client can ingest their own documents (policies, listings, HOA/MLS rules) and ask grounded questions that return concise answers with citations.

Features

• Ingestion via JSON or file upload (pdf/txt/md) with automatic chunking.
• Per-tenant TF‑IDF indexing on disk under data/processed/{client_id}.
• Deterministic dummy LLM for local development and tests (no API key required).
• Optional local Transformers model when RAG_MODEL_NAME is set (e.g., google/flan-t5-small).
• REST endpoints:
  • GET /healthz
  • POST /v1/ingest-json
  • POST /v1/ingest-files
  • POST /v1/ask

Quickstart

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python main.py  # starts the API on :8000

Open the colorful UI at:

http://localhost:8000/

To use a local Transformers model:

pip install transformers torch  # run once
RAG_MODEL_NAME=google/flan-t5-small python main.py

Running Tests

pytest