TravoAI

You log in with Google and fill out a short, Typeform-style questionnaire about the trip: where, how many days, the budget, how many people, and the group's interests.

TravoAI turns that into a prompt for OpenAI's text-davinci model, which returns a full itinerary as JSON, balanced against the group's interests, budget, and size.

The itinerary renders day by day. From there you can grab a shareable link to send a friend, export every event straight to Google Calendar, or ask the built-in chatbot anything about the trip.

Deciding what actually belonged in the MVP was the first hurdle. We spent real time brainstorming features only to find some weren't feasible in the time we had, or didn't fit the product.

Working with text-davinci was the next one. We had to tune our prompts to keep itinerary generation fast. We also started on React, then realized plain HTML, CSS, and JavaScript would simplify passing data between front and back end. The switch dropped us from two servers to one, but cost us some time refactoring. It was also my first time with EJS, so rendering backend data took some learning.

Despite the time crunch and the last-minute changes, we shipped, on not a lot of sleep. Building it together was the fun part, and we're proud of what we made.

With more time, a business model: a destination like Science World could surface as a top suggestion for people planning a Vancouver trip, and pay a small cut for each visitor TravoAI sends their way.

We also wanted bus routes, costs, reviews, and images in the itinerary, plus weather, eco-friendly stays, a booking system, a feedback loop, and a way to connect solo travelers heading to the same place.