NYC POI Concierge 🗽

Prestige-first, context-aware restaurant discovery for NYC. Built for the MongoDB x Tavily x LastMile AI Hackathon (Nov 2025).

🚀 Overview

A mobile concierge that moves beyond generic recommendations. We use Tavily to curate Michelin-starred and award-winning spots, MongoDB Atlas for geospatial + vector search, and LastMile AI's MCP to power a context-aware chat interface.

🎥 Demo Video

Watch the full walkthrough:

🔗 Watch on YouTube

🛠️ Tech Stack

• Database: MongoDB Atlas (Geospatial + Vector Search)
• AI/Search: Tavily API (Real-time enrichment), OpenAI GPT-4o
• Backend: LastMile AI MCP Server (Python)
• Frontend: Expo React Native (iOS)

📂 Structure

• backend/mcp-server/: MCP server implementation
• frontend/expo-app/: React Native mobile app
• scripts/: Data pipelines, verification, and ops tools
• data/: Curated and raw POI datasets

⚡ Quick Start

1. Backend (Local)

cd backend/mcp-server
pip install -r requirements.txt
python3 http_server.py
# Runs on http://localhost:8000

2. Frontend (Mobile)

cd frontend/expo-app
npm install
npx expo start
# Press 'i' for iOS simulator

☁️ Deployment

• MCP Cloud: Deployed and accessible via MCP Client.
• Data: 130+ curated POIs in MongoDB Atlas.

🧪 Testing

Run the verification suite:

python3 scripts/verification/check_fine_dining.py