DreamTrail AI

Plan the journey your heart remembers

Comment

Inspiration

We noticed that most travel apps ask you where you want to go first. But most people don't start there. They start with a feeling — "I want to feel calm," or "I want an adventure," or "I want to disconnect from my phone for a week." We wanted to build something that starts from that feeling instead of a destination search box. So we built DreamTrail AI to turn an emotional wish into a real, usable trip plan.

What it does

You describe the trip you want in plain words, plus basics like origin, days, budget, and pace. The app gives you a destination with a reason why, a full day-by-day itinerary, a budget breakdown with savings tips, and a short pre-trip story in a style you pick. You can ask for changes in plain language, and after the trip, turn your own photos and notes into a post-trip story.

How we built it

Frontend is React + TypeScript + Vite with Tailwind. Backend is Python + FastAPI + Pydantic. Google Gemini generates the actual plan, returning strict JSON that we validate before it reaches the user. Two of us split the work — one on backend/AI, one on frontend — after agreeing on the data contract first. We used Codex and GPT a lot for planning and scaffolding.

Challenges we ran into

Keeping AI output reliable and in the right shape was the biggest one. Small bugs (like a mismatched function signature) once broke every "revise" request. We also had to resize and cap photo uploads so the post-trip story feature didn't get slow or expensive. And we had to be careful to label every cost as an estimate, not a real price.

Accomplishments that we're proud of

The full flow works end to end: prompt → itinerary → budget → story → revise → post-trip story. The budget feature gives real trade-off suggestions when you're over budget. We also shipped the post-trip photo story, which started as a stretch goal.

What we learned

Forcing structured AI output and validating it strictly is much safer than parsing free text. A clear shared contract between frontend and backend saved us a lot of time. Small mismatches are easy to miss, so testing each endpoint properly matters more than it seems.

What's next for DreamTrail AI

Real accounts with persistent saved trips, accessibility/accommodation preferences in the form, live AI stage progress instead of a fixed loader, real pricing data, maps and weather, and support for group trip planning.

Built With

Share this project:

Updates