PropQuora — AI-Powered Real Estate Platform
Production-grade real estate SaaS with conversational AI search, PostGIS map queries, and Gemini-grounded responses
Overview
PropQuora is a final-year SaaS platform combining a conversational AI search interface with a map-based property browser, targeting the Pakistani real estate market (Islamabad). A two-pass Gemini pipeline parses natural language into structured filters, runs real PostGIS queries, then generates grounded responses — no hallucination. The map layer uses Mapbox GL JS with client-side clustering and pg_tileserv serving vector tiles directly from PostGIS materialized views.
Supports four roles (platform admin, agency admin, agent, consumer) with full JWT auth, listing CRUD with status workflows, agency applications, and statistical price prediction built from the Zameen.com dataset. Hybrid semantic + lexical search (pgvector cosine + pg_trgm trigram) resolves ambiguous Pakistani location names like 'F-7', 'DHA', and 'NUST'.
Deployed via Docker Compose + Nginx with Alembic migrations running on startup.
Key Features
Conversational AI search — a two-pass Gemini pipeline parses natural language into real PostGIS queries, then generates grounded, hallucination-free answers
Radius and area map search on Mapbox GL with live price-marker clustering across dark, streets, and satellite styles
Hybrid semantic + lexical search (pgvector cosine + pg_trgm) that resolves ambiguous location names like 'F-7', 'DHA', and 'NUST'
Vector tiles served directly from PostGIS materialized views via pg_tileserv
Four roles with JWT auth: platform admin, agency admin, agent, and consumer
Side-by-side listing comparison and nearby-facilities POI analysis
Statistical price prediction trained on the Zameen.com dataset
Docker Compose + Nginx deployment with Alembic migrations on startup
Project Gallery
