AI Travel Assistant

Login Page
First Chat
Home Page

The Voyago Story: Building the Future of AI Travel

Inspiration

Planning a trip today is surprisingly fragmented. Travelers constantly switch between Google Maps, travel blogs, restaurant reviews, weather websites, booking platforms, and itinerary apps just to organize a single vacation.

We wanted to eliminate this context switching by creating Voyago—an AI-powered travel companion that combines conversation, intelligent planning, and interactive visualization into one seamless experience.

Rather than forcing users to search for information manually, Voyago understands natural language, performs the required actions autonomously, and presents the results through an interactive map and itinerary.

⸻

What it does

Voyago is an intelligent AI travel planner that transforms natural language into complete travel experiences.

Users can simply ask questions like:

“Plan a 3-day trip from Calgary to Banff.”
“Find hidden photo spots along my route.”
“Show me Italian restaurants and cafés on the way.”

Voyago automatically:

Generates optimized driving routes
Finds attractions, restaurants, cafés, viewpoints, and other places of interest
Displays everything on an interactive map
Builds a structured multi-day itinerary
Maintains conversational memory so trips evolve naturally as users continue chatting

Instead of returning only text, Voyago synchronizes every recommendation with a live map and timeline, creating an experience that is both conversational and visual.

⸻

How we built it

Voyago is built using a modern cloud-native AI architecture.

Frontend

Next.js (App Router) for a fast and scalable React application
Tailwind CSS for a responsive glassmorphism interface
Leaflet + React Leaflet for real-time interactive maps
Dynamic component loading using next/dynamic to support client-side map rendering

Backend

FastAPI powering REST APIs and WebSocket communication
PostgreSQL for persistent chat history and user data
JWT Authentication for secure user sessions

AI Agent

Our travel agent follows a tool-using architecture.

Instead of relying purely on an LLM, the AI decides which specialized tools to invoke, including:

Route generation
Place search
Route-aware place discovery
Interactive itinerary generation

These tools retrieve structured information before the LLM generates the final response, reducing hallucinations while producing grounded travel recommendations.

External Services

Voyago integrates with:

Mistral AI for natural language reasoning
Mapbox Geocoding API for location resolution
OSRM for route generation
OpenStreetMap Overpass API for discovering nearby attractions and businesses

Observability

Every AI interaction is monitored using Langfuse, allowing us to trace prompts, tool executions, latency, and model outputs for debugging and evaluation.

⸻

Challenges we ran into

One of our biggest technical challenges was building a reliable AI agent that could coordinate multiple external services.

We encountered:

Rate limiting from OpenStreetMap during large searches
Ambiguous city names causing failed route generation
Synchronizing map updates with AI responses in real time
Parsing structured itinerary data generated by the LLM
Managing WebSocket communication while keeping chat history persistent

On the frontend, integrating Leaflet with Next.js required dynamic loading because browser-specific APIs are unavailable during server-side rendering.

⸻

Accomplishments that we’re proud of

We’re especially proud that Voyago goes beyond being “just another chatbot.”

Some highlights include:

A tool-using AI agent capable of planning complete trips
Real-time synchronized map updates driven by AI decisions
Interactive route visualization with searchable points of interest
Automatic generation of structured multi-day itineraries
Persistent chat history with context-aware conversations
Production-ready cloud deployment with monitoring and observability

Most importantly, we created an experience where users can simply describe their ideal trip and watch it come to life visually.

⸻

What we learned

Building Voyago reinforced several important lessons.

We learned that modern AI applications are far more powerful when LLMs are combined with external tools rather than relying solely on language generation.

We also gained experience designing AI agent workflows, integrating mapping services, handling asynchronous WebSocket communication, and building scalable cloud-native applications with proper observability.

⸻

What’s next for Voyago

Our vision is to evolve Voyago into a fully autonomous travel assistant.

Future enhancements include: