RentAI MVP

An end-to-end demo that mixes classic property filtering with AI ranking and AI powered market insights. The repo bundles a NestJS + Prisma backend, a Vite/React frontend, a Dockerized Postgres database, and Python helpers for LLM ranking and POI precompute.

Our problem

We all know how tedious finding the perfect new home in a big city can be. From having to doomscroll through thousands of properties, to fake properties that don't exist, to duplicate announcements and to having to chat with tens of owners/agents. This is where we come in.

Our vision

We provide a filtered, fully personalized and automated experience for our clients. Log into our platform, create an account, and desribe what you are looking for. From your profile description, we run an analysis on the available properties and use our AI-powered reasoning system to find you the properties best tailored for your needs.

Features

Current :

• Best-match real estate AI-powered lookup
• AI-powered property price comparison to surrounding similar offers
• Fully automated chat with agent/owner
• Smart-filter that determines individual benefits for each property

Upcoming :

• Address validation for each listing
• AI powered description of each listing based on the listed images

NOTE : In order to test the functionality locally, on your device you will need Gemini and Grok API keys

Quick start

git clone https://github.com/nchristiann/rentAI.git
cd rentAI/project

# 1) start Postgres
cd infrastructure
docker compose up -d

# 2) configure backend env & deps
cd ../backend
cp .env.example .env      # add your secrets manually afterwards
npm install
npx prisma db push
npx prisma generate

# 3) run backend
npm run start:dev

# 4) configure frontend env & deps
cd ../frontend
cp .env.example .env.local   # optional, only if you need custom base URL
npm install

# 5) run frontend
npm run dev

Backend runs on http://localhost:3000, the Vite dev server defaults to http://localhost:5173 with a proxy so the React app can call /api/... without CORS issues.

Backend stack

Layer	Details
Framework	NestJS with feature modules (listings, preferences, ai, grok, auth, etc.)
ORM	Prisma connected to Postgres via DATABASE_URL
Database	Postgres (Docker compose file in infrastructure/docker-compose.yml)
AI helpers	Python scripts in backend/src/ai (LLM reranker, POI precompute) + the new GrokClient

Important commands (run inside backend/):

npm install              # install dependencies
npx prisma db push       # sync schema to local DB
npx prisma generate      # regenerate Prisma client
npm run start:dev        # start Nest in watch mode

AI reranking + Grok

The /api/ai/rerank endpoint now auto-detects which provider to use:

1. If RERANK_LLM=grok or GROK_API_KEY is present, the Nest service calls Grok directly via the new GrokClient.
2. Otherwise it falls back to Gemini (same payload/prompt, still available if you keep the Python dependency installed).

To enable Grok everywhere, add to backend/.env:

GROK_API_KEY="sk-your-key"
GROK_MODEL="grok-beta"         # optional override
GROK_API_URL="https://api.x.ai/v1/chat/completions"

Restart the backend after editing .env.

Frontend stack

• Vite + React single-page app (frontend/src).
• Context providers for auth/preferences, a Listings.jsx page with algorithmic filters, two-step AI flow, and Grok badges.
• Leaflet map overlay (components/MapOverlay.jsx) for spatial exploration.

Key commands (inside frontend/):

npm install
npm run dev        # start Vite dev server with proxy
npm run build      # production build (requires Node 22.12+ per .nvmrc)

To point the UI at a different backend origin, edit frontend/.env.example (or .env.local) and set VITE_API_BASE_URL.

Database + seed data

1. docker compose up -d inside infrastructure/ starts Postgres on port 5432 with deterministic seed SQL scripts (listings, sellers, etc.).
2. Optional extras: run infrastructure/init/add_images_unsplash.sql or seed_listings_extra.sql to load demo imagery and neighborhoods.
3. For POI metrics, execute python backend/src/ai/precompute_worker_v2.py --force once Postgres is populated.

Use Prisma Studio (npx prisma studio) or plain psql to inspect data.

Troubleshooting cheatsheet

Symptom	Likely cause	Fix
Frontend overlay says "[AI Fallback]"	/api/ai/rerank failed (missing key, timeout)	Inspect backend logs; ensure Grok/Gemini env vars are set
Cannot find module '@google/genai'	Backend deps not installed	npm install in backend/
npm run build warns about Node version	Node < 22.12.0	nvm install 22.12.0 && nvm use
404 /api/listings in dev	Proxy misconfigured	Run frontend via npm run dev (Vite proxy) or set VITE_API_BASE_URL
Empty listings table	Seeds not applied	docker compose exec postgres psql ... -f infrastructure/init/seed_listings.sql

Contributing

1. Create a feature branch, keep commits scoped.
2. Run lint/build/tests for both backend and frontend where relevant.
3. Ensure new environment variables have sane defaults and are documented here.
4. Open a PR with a short demo note or screenshots.

---