RealtyAssistant AI Agent

Production-Ready AI Voice & Chat Agent for Real Estate Lead Qualification

🏠 Overview

RealtyAssistant is an advanced AI system designed to qualify real estate leads through natural voice conversations and intelligent chat interactions. Unlike simple chatbots, it features a fail-safe architecture, dynamic accent understanding, and real-time property search capabilities.

It is designed to run on a cheap VPS (Ubuntu) or local Windows machine with a single "One-Click" setup command.

Key Features

• 🎙️ Natural Voice Interface:
  • Handles Indian accents (Noida, Gurugram, etc.) via fuzzy matching.
  • "Rich Input" processing: "I want a 3 BHK in Noida under 50 lakhs" is understood instantly.
  • Voice-to-Text via local Whisper engine.
• 🛡️ Fail-Safe Leads Database:
  • Auto-Healing: Automatically detects and repairs database corruption.
  • Zero-Loss: Falls back to in-memory processing if disk fails, ensuring leads are captured in logs.
  • Structured Storage: SQLite database tracking every conversation detail.
• 🔍 Live Intelligence:
  • Scrapes realtyassistant.in in real-time to show actual property listings.
  • Hybrid LLM (Ollama + Gemini Fallback) for robust understanding.
• 🚀 Production Ready:
  • Runs on Port 20000 (configurable).
  • Includes Nginx configs and systemd service files.
  • One-click deployment scripts for Windows and Linux.

🚀 Quick Start (One-Click)

Windows (Development)

Double-click run_project.bat or run in terminal:

run_project.bat

This script automatically:

1. ✅ Checks for Python 3.10+.
2. ✅ Creates a virtual environment.
3. ✅ Installs all dependencies.
4. ✅ Frees up Port 20000 if occupied.
5. ✅ Starts the Server.

Access: http://localhost:20000/demo

Linux (Production / VM)

sudo bash run_project.sh

This script automatically:

1. ✅ Installs System Dependencies (Python, Nginx, etc.).
2. ✅ Sets up the AI Environment.
3. ✅ Configures Nginx Reverse Proxy.
4. ✅ Installs it as a background Systemd Service.

Access: https://<your-ip>/task1/demo

🏗️ Architecture

For a deep dive into the system design, fail-safe mechanisms, and voice pipeline, see ARCHITECTURE.md.

┌──────────────┐      ┌─────────────┐      ┌───────────────┐
│  Voice/Chat  │      │  Analysis   │      │  Persistence  │
│  Interface   │ ───> │   Engine    │ ───> │   Database    │
└──────────────┘      └──────┬──────┘      └───────┬───────┘
                             │                     │
                      ┌──────▼──────┐      ┌───────▼───────┐
                      │  Property   │      │   Fail-Safe   │
                      │  Searcher   │      │   Mechanism   │
                      └─────────────┘      └───────────────┘

🔧 Configuration

The application is configured via .env. A default one is created automatically if missing.

Variable	Default	Description
PORT	20000	Application Port
GEMINI_API_KEY	-	Required for AI fallback
DATABASE_URL	sqlite:///data/leads.db	Main database path
OLLAMA_BASE_URL	http://localhost:11434	Local LLM URL
VOICE_PROVIDER	twilio	twilio or vapi

📡 API Endpoints

Core

• GET /demo - Chat Widget Interface
• GET /voice - Voice Testing Interface
• GET /leads - Admin Dashboard for Leads

API

• POST /api/qualify - Submit lead data
• POST /api/leads - Create lead entry
• GET /api/leads - Fetch lead history
• POST /api/voice/process - Handle audio chunk

🧪 Testing Voice

1. Navigate to http://localhost:20000/voice.
2. Click the Microphone button.
3. Speak naturally. Try saying:

   "Hi, I am looking for a 3 BHK apartment in Noida Extension."

4. The system will:
   • Recognize your speech.
   • Extract "3 BHK", "Apartment", "Noida Extension".
   • Skip redundant questions and verify your details.
   • Search for properties.

📄 License

MIT License. Built with ❤️ by dmj.one.