Inspiration
Personally wanted to explore the intersection between AI and Blockchain and began exploring the AI companies that currently exists. While setting up my GitHub, saw the side-jobs that I did at my internship(turtle.xyz) and the idea came to make it easier to get paid with open-source projects
The goal of this project was to incentivize/increase open-source project engagement and hopefully provide more opportunities for students to earn money / experience on the side.
What it does
It's a GitHub App that helps organize and standardize the process of managing the reviews and payment to PRs.
People who want to use this service can just install it as a GitHub App, add required secret keys on their GitHub Repo, and merge the pre-written branch that includes Workflow Scripts, managing/sending the automated transactions from the GitHub bot.
For people who want to contribute to a repo simply just needs to add their Sui wallet address in their initial PR post (If they wish to change it, they can comment the new wallet address, in which the bot will pick up the hidden html line).
Once the PR is deemed safe/error-free by CodeRabbit, the bot will then use a OpenAI to review the feedback by CodeRabbit and set a price to the PR, in which the person who contributed can either accept, or contact the admin to readjust the price.
If the contributor accepts, Github will then accept the transaction and send the valued Sui from their Treasury wallet to their address.
How I built it
I used a GitHub App built with Node.js + Probot to listen for pull-request events. When CodeRabbit Reviews code, have it listening to when it gives feedback, in which it determines wether or not its rejected/accepted and whether or not to continue with valuing the PR or not. When a PR is merged, the app triggers a GitHub Actions workflow. The workflow installs the Sui CLI and loads a temporary keystore using repo secrets. It signs a transaction from a treasury wallet and transfers SUI to the contributor. Everything runs inside GitHub Actions, so no one needs to run code locally.
Challenges I ran into
Since this was a solo project, found it really hard to manage all the aspects of the app and even trying to ship the MVP.
Also, since this was my first time creating a GitHub App, creating scripts for Workflows, and building on top of Sui, it got pretty intense trying to juggle reading the docs and learning about what/how to use it and merge it together.
Speaking of merging together, the actual Sui section of this project needed to all be on the Github workflow (CLI), in which, figuring out how to Install the Sui CLI in the workflow, creating the client (since every time the workflow is compartmentalized by - name:, behind the scenes it creates a new terminal, causing the client.yaml file to be empty), and importing the wallet / sending the transaction, was a huge headache.
Accomplishments that I'm proud of
I'm overall pretty proud that I was able to make a project that works considering all the factors above.
What I learned
Not only did I learn how to make a GitHub App, workflow, Sui CLI, and having different AI communicate with each other (not really), I think the most important lesson is to find a partner next time to share this headache with.
What's next for GitSui
Features that I wanted to add but couldn't due to time:
- Multi-sig human/ai verification on PR
- Make contributor commits/contributions into a NFT in which could act as a trophy/evidence to their contributions
- More in-depth feedback from Open-AI instead of just scoring
Demo will be here: https://drive.google.com/drive/folders/1nIkAEt2sGl7fNpLr2xxBZllv7t_vrg1j?usp=sharing
Built With
- bash
- javascript
- yaml



Log in or sign up for Devpost to join the conversation.