AI Document Assistant for Real Estate Brokerage with RAG Pipeline

Overview

We built an AI-powered case management system for a real estate brokerage, with a React chat frontend. Agents query client records, upload PDFs and photos, and summarize property contracts via RAG — including multi-turn conversations extracting specific contract clauses like closing costs and seller disclosures. Built across 3 n8n workflows: Airtable as CRM, Google Drive as document storage with per-client folders, and a RAG pipeline using Supabase pgvector and OpenAI embeddings. Our PDF-replace logic deletes stale vectors before re-indexing. Weekly HTML report per agent, delivered via Gmail every Monday.

Technical Highlights

• 1Folder-per-client structure in Google Drive — auto-created on first upload
• 2Duplicate file check prevents version conflicts before upload
• 3Stale vector deletion before re-embedding prevents outdated chunks in RAG
• 4Weekly HTML report with case breakdown sent to each agent individually
• 5React chat interface connected to n8n via webhook