Code Review AI

We don’t teach how to do code reviews. However, projects rely on them?

Comment

Inspiration

We are generally not taught how to do code reviews, but this is an essential part in every programmers day to day task. Thus, creating a platform/tool which helps them learn how to do code reviews seems obvious. And this is what I did!

What it does

The project takes an existing public github repo, downloads some parts of it and introduces new code review opportunities which you have to find out and explain WHY these changes are wrong.

How I built it

I built it using NextJS, Tailwind, Gemini API and MongoDB Atlas so that I can ship fast :)

Challenges I ran into

One of the biggest challenges was working with Gemini API: managing rate limits, controlling the scope of AI changes and ensuring that the changes are still challenging enough for the developer to actually think about what is wrong, plus, defining the right balance for the scope as a solo developer was also difficult and extremely exhausting.

Accomplishments that I am proud of

I am proud of creating an experience that works on real user-provided repositories instead of synthetic examples. (And for sure finishing the project alone)

What I learned

I learned that managing an AI API is pretty troublesome for testing and that managing projects in MongoDB is not just adding a list of strings.

What's next

I plan to incorporate the feature in Github so that the users can directly get some practice on a platform they already use.

Built With

Share this project:

Updates