PHS Hackathon - Flight Ready

Inspiration

We have all been there: after hours waiting in line at the airport, you finally make it to security, just to get stopped all of a sudden because your water bottle was too big. Our group decided we wanted to do something about it. Therefore, we decided that we wanted to catch those problems at home, not at the airport. With the theme "Into the Skies", we built a packing assistant that gets you ready before you leave for the airport.

What it does

Our program lets you take a photo of the things you're packing. An image-identification AI (Google Gemini) then identifies each item and checks it against a list of TSA rules.

How we built it

Frontend: plain HTML, CSS and JavaScript with a cartoon airplane theme. It’s built for phones and works with a keyboard and screen readers. Backend: a Flask server with a JSON API for scanning, saving items and the final check. AI: the Gemini API reads the photo and returns each item’s name, category, quantity and charging needs as structured JSON. TSA rules list: a rules file has the final say over the AI’s guess, so the safety check doesn’t depend on the AI alone. The rules also differ by bag, since scissors are a warning in a carry-on but fine in a suitcase. Teamwork: three of us split the frontend and backend and worked together through GitHub.

Challenges we ran into

Connecting the frontend + backend: This included serving ports and keeping both sides agreeing on the same data format Making AI output reliable: We had to clean up the AI’s answers (missing fields, wrong types) before the app could trust them. Keeping the TSA check trustworthy: We made our own list that is TSA-compliant that overrides the AI's decision-making. Time: We only had around 24 hours so we had to avoid burnout, take breaks in between, and made sure we had the key features Deployment: We had to deploy on fly.io without any errors, which was a major obstacle

Accomplishments that we're proud of

-A full flow that works end to end, with a photo, scan, TSA check, bag, and final check. -A TSA check that goes beyond one AI answer, with rules that differ by bag. -A friendly design with an accessible layout that works on a phone. -Creating something that has a practical use. -Fully deployed and functional application -Getting three people to work on and complete a project in only 24 hours.

What we learned

-How to use and manage GitHub branches -How to send a photo from a browser to a Flask server and then to an AI model. -How to agree on a data format early so frontend and backend work in parallel. -How to turn an AI’s answer into data an app can use safely.