This is a submission for the Built with Google Gemini: Writing Challenge
What I Built with Google Gemini
I created a project called Gemini Mechanic. It is a smart assistant that helps people fix broken hardware and electronics. Many people feel lost when a gadget stops working. My web app lets users take a photo of the problem or type a description to get help right away.
Google Gemini is the brain of this project. It looks at the photos to figure out what is wrong. Then it gives the user a list of repair steps and tells them what parts they need to buy. This makes it much easier for regular people to repair their own things.
Demo
I previewed it and realised the embedded version doesn’t allow the camera to be used so you can just access the live link here:
In my demo, I used my broken iPhone 8 as an example. The app was able to recognize the phone and see that the screen was detached. It provided a full repair guide. It included steps to turn off the power and remove the ribbon cables safely.
What I Learned
One of the biggest lessons I learned was about mobile design. Most people are not at a desk when they are trying to fix something. They are usually holding their phone while they work. I spent a lot of time making the app mobile responsive so it looks good on small screens.
I also learned how to write better instructions for the AI. For repairs, safety is the most important part. I had to make sure Gemini always reminds users to do things like disconnect the battery before they start working.
Google Gemini Feedback
Working with Gemini was a good experience, but I did face some challenges:
Picking the right model: I spent a lot of time testing different Gemini models. I needed one that was smart enough to recognize small parts but fast enough to work on a phone. I don’t know why whenever I try to use Gemini as the ai for any of my projects, I always run into errors. But with the project I was pushed to figure it out, took me a couple of hours till I found the best model to run it with.
Mobile speed: Making sure the app stayed fast while the AI was thinking was tricky. I had to optimize the code so the user experience stayed smooth on mobile devices.
Open Source
I want to make this project open source so everyone can use it. If you have ideas for new features or want to add more repair data, please feel free to contribute to the GitHub repo. I would love to see what we can add to make it better.
Top comments (5)
Massive Project Maame, you're doing a phenomenal job. And the UI ? Looks slick and nice imo :D
Thank you Luftie for the kind words :)
I add LangChain, LangGraph, LangSmith,and MCP server give more support to the gemini model for my projects
Thanks for the add up Benjamin! I will definitely try incorporating that in this project later on because I always struggle with using Gemini honestly. Massive help thanks
very good use of vision ai!