About

Built by a traveller,
powered by AI

TravelBuddy is a personal project combining a passion for travel with an interest in making AI tools that are actually useful.

Paolo Cremonese

Paolo Cremonese

I built TravelBuddy as a side project to explore what AI agents can do for real-world planning tasks. I love travelling and thought itinerary generation was the perfect testbed — it needs live data (weather), structured output, multi-step reasoning, and a decent UI.


How it works

Each itinerary runs through a pipeline of five AI calls in sequence. Every step is a separate prompt sent to the selected model.

1
🌤 Weather check

Fetches live weather from Open-Meteo for the destination coordinates, then asks the AI to summarise it and flag any travel advisories (too hot, rainy season, etc.).

2
✍️ First draft

The AI plays a local tour guide. Given your city, interests, trip duration, pace, and the weather summary, it writes a detailed day-by-day itinerary.

3
📍 Extract stops

A second prompt parses the draft and extracts a clean semicolon-separated list of stops — each with name, city, and country.

4
✅ Validate stops

The stop list is sent back to the AI for a sanity check: are they real places? In the right city? In a logical walking order? Duplicates removed.

5
🎨 Format itinerary

The draft and validated stops are combined into the final HTML output, including the per-stop 'More info' deep-link URLs and useful tips.

🗺
🗺 Interactive map

The stop list is passed to the Google Maps Embed API to render a live directions map with the full route — no extra AI call needed.


Tech stack

Backend
Python / Flask
AI inference
Groq (LPU) — Llama 3.3 70B
Maps
Google Maps Embed API
Weather
Open-Meteo (free, no key)
Geocoding
Nominatim / Google Geocoding
Hosting
Hugging Face Spaces

Ready to plan your trip?

No sign-up. No cost. Just tell us where you want to go.

Create an Itinerary