Your AI-powered travel planning assistant - transforming fuzzy prompts into concrete, bookable itineraries
WanderGenie is an intelligent travel planning assistant that uses multi-agent AI to transform natural language travel desires into complete, personalized itineraries. Simply tell us your travel plans, and we'll handle the rest.
Travel planning today means juggling multiple tabs: Google for attractions, Maps for routes, spreadsheets for schedules, booking sites for hotels and flights. It's fragmented, time-consuming, and overwhelming.
One conversational interface. Complete trip planning.
Type: "5 days in NYC from Buffalo, Dec 20-25, with a teen, love views & pizza, avoid long lines"
Get:
- ✅ Day-by-day personalized itinerary with times & travel estimates
- ✅ Interactive map with location pins
- ✅ Calendar integration (Google Calendar or .ics export)
- ✅ Pre-filled booking links (flights, hotels, attractions)
- ✅ Conversational edits: "Swap Day 3 afternoon for MoMA" → updates instantly
See WanderGenie in action!
[4-minute demo showcasing trip planning from natural language input to complete itinerary with maps, timeline, and booking links]
Powered by LangGraph with three specialized AI agents:
- Planner Agent: Interprets your travel intent and preferences
- Researcher Agent: Discovers and enriches points of interest using VectorDB + GraphDB
- Packager-Executor Agent: Builds optimized day-by-day schedules with smart routing
- VectorDB (pgvector): Retrieval-augmented generation for booking tips, local insights, and constraints
- GraphDB (Neo4j): Relationship-aware clustering for neighborhoods, similar attractions, and ticket vendors
- Smart Caching: Offline fallbacks ensure demos never break
Make changes naturally:
- "Add Joe's Pizza to Day 2 lunch"
- "Move the Statue of Liberty to Day 1 morning"
- "Replace outdoor activities with museums on Day 3"
The entire itinerary, map, and calendar update automatically.
- Interactive Maps: Mapbox/Leaflet integration with pins, routes, and day-by-day clustering
- Timeline View: Clean, scannable schedule with travel time estimates
- Progress Indicators: Watch agents work in real-time with status chips
- Google Flights: Pre-filled origin, destination, and dates
- Hotels: Booking.com with accurate check-in/check-out dates and guest count
- Attractions: Direct links to official booking sites (Statue Cruises, museum tickets, etc.)
- LLM-Powered POI Generation: Works for any city worldwide, not just pre-seeded locations
- Intelligent Caching: First-time generated POIs are saved to the database for faster future requests
- State Name Inference: "Florida" automatically converts to "Miami, FL" for better results
- Growing Knowledge Base: Every new city query enriches the system's POI database
- Backend: Python + FastAPI + LangGraph
- LLM: AWS Bedrock (Claude) with OpenAI fallback
- Memory Layer:
- VectorDB: Supabase pgvector
- GraphDB: Neo4j Aura Free
- State Store: Supabase (PostgreSQL)
- Frontend: React + TypeScript + Tailwind CSS + Mapbox GL
- Deployment: Cloud-ready (Docker support included)
User Prompt
↓
Planner Agent → Parse intent (city, dates, preferences)
↓
Researcher Agent → Find POIs
├─ Check VectorDB cache first
├─ Query OpenTripMap + GraphDB
└─ LLM Fallback (if needed) → Save to cache
↓
Packager-Executor → Build schedule + Generate map/calendar/links
↓
Validator → Schema check + Auto-patch issues
↓
Complete trip.json → UI Update
See ARCHITECTURE.md for detailed diagrams and data flows.
- Python 3.10+
- Node.js 18+
- AWS Account (Bedrock access)
- Supabase Account (or local PostgreSQL + pgvector)
- Neo4j Aura Account (free tier)
- Clone the repository
git clone https://github.com/PatilPrajakta14/WanderGenie-ai-travel-assistant.git
cd WanderGenie-ai-travel-assistant- Backend Setup
cd backend
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -r requirements.txt- Frontend Setup
cd Frontend/wandergenie
npm install- Configure Environment Variables
Create .env in the root directory:
# LLM
AWS_ACCESS_KEY_ID=your_aws_key
AWS_SECRET_ACCESS_KEY=your_aws_secret
OPENAI_API_KEY=your_openai_key
# Databases
SUPABASE_URL=your_supabase_url
SUPABASE_KEY=your_supabase_anon_key
NEO4J_URI=neo4j+s://your-instance.databases.neo4j.io
NEO4J_USER=neo4j
NEO4J_PASSWORD=your_password
# APIs
OPENTRIPMAP_API_KEY=your_opentripmap_key
GOOGLE_CLIENT_ID=your_google_client_id # Optional
GOOGLE_CLIENT_SECRET=your_google_secret # Optional- Seed Databases (Optional - system will auto-generate POIs via LLM)
# Seed VectorDB (pre-load NYC data)
python backend/scripts/seed_vectordb.py
# Seed GraphDB (pre-load NYC relationships)
python backend/scripts/seed_graphdb.py- Run Development Servers
# Terminal 1: Backend
cd backend
python3 -m uvicorn backend.main:app --reload --port 8000
# Terminal 2: Frontend
cd Frontend/wandergenie
npm startVisit http://localhost:3000 to see WanderGenie in action! 🎉
WanderGenie-ai-travel-assistant/
├── Frontend/
│ └── wandergenie/ # React application (TypeScript)
│ ├── src/
│ │ ├── components/ # React components
│ │ ├── hooks/ # Custom hooks
│ │ ├── pages/ # Page components
│ │ ├── services/ # API client
│ │ └── utils/ # Utilities
│ └── public/ # Static assets
├── backend/ # FastAPI + LangGraph
│ ├── agents/ # Planner, Researcher, Packager
│ ├── tools/ # POI search, distance, links, etc.
│ ├── memory/ # VectorDB + GraphDB clients
│ ├── routes/ # API endpoints
│ ├── schemas/ # Pydantic models
│ └── scripts/ # Database seeding scripts
├── data/ # Seed data & fallback caches
│ ├── nyc_pois.json
│ ├── poi_facts.csv
│ └── neo4j_seed.cypher
├── docs/ # Documentation
│ ├── ARCHITECTURE.md
│ ├── API.md
│ └── VECTORDB_IMPLEMENTATION.md
└── tests/ # Test suite
UB Hacking 2025 - Team WanderGenie
- Sweta Sahu - LLM/Agent Lead
- Gautam Arora - Backend/API Lead
- Arpit Sharma - Frontend Lead
- Prajakta Patil - DevOps/Data Lead
GitHub: https://github.com/PatilPrajakta14/WanderGenie-ai-travel-assistant
Made with ❤️ at UB Hacking 2025 | November 8-9, 2025