Reparo

Inspiration

In an era of rising living costs and "planned obsolescence," Canadians are feeling the squeeze. When something breaks, our first instinct is often to throw it away because finding the right part or knowing how to fix it feels overwhelming. Reparo was born from the idea that the "Right to Repair" should be accessible to everyone. We want to help Canadians save their hard-earned money and keep usable items out of landfills.

What it does

With a picture of a broken object (e.g. broken chair leg, hole in clothing), our computer vision model identifies the specific component that has failed. Gemini then analyzes the damage, generates a step-by-step DIY repair guide with estimated cost, difficulty, and repairability information available, and automatically populates a shopping cart with the exact replacement parts and tools needed from Shopify/other online retailers

How we built it

We used Swift and Reactiv to develop it as an App Clip, which allows users to start repairing instantly without a full app installation. Our backend uses Python which orchestrates the image processing and communicates with the Gemini API to transform visual data into actionable instructions. As well, we used SerpAPI to find primarily Shopify stores selling required items, and then used Shopify JSON and public endpoints to get item data.

Challenges we ran into

One challenge we ran into was figuring out how to make our app easy to buy real parts from. Eventually, we realized we could utilize Shopify stores' uniform JSON structure to make parts purchaseable for users directly. We also navigated the steep learning curve of integrating Reactiv with Swift for a seamless mobile e-commerce experience.

Accomplishments that we're proud of

Using Swift and creating a mobile app that helps users be sustainable and save money.
Seamless Integration: Getting a vision model to successfully "hand off" data to generate a real-world shopping cart.
The UX of Reactiv: Learning how Reactiv can streamline the checkout flow in a mobile environment opened our eyes to the future of "instant-on" e-commerce.

What we learned

We learned that the hardest part of AI isn't the model itself—it's the data pipeline. Structuring visual metadata so that an AI can provide safe and accurate repair advice is a delicate balance. We also gained a deep appreciation for the App Clip ecosystem; reducing friction is the best way to encourage sustainable habits.

What's next for Reparo

AR Assistance: We want to implement AR overlays so you don't just read the instructions—you see the digital "ghost" of the part fitting into place.
Partnering with Canadian retailers like Home Hardware or Rona to allow for "click-and-collect" repairs.