We at Mibex recently held another hackathon, and this time the theme was focused on the potential of AI to improve our existing apps. At the event, our team collaborated to investigate AI potential and devise innovative solutions. The outcome? Brand new applications and a bunch of learnings.
The Hackathon
The hackathon took place over three days in April, and it was an opportunity for us to learn more about the latest developments in AI and how we could incorporate them into our work.
We were excited to work with large language models like ChatGPT and explore how they could help us automate tasks in regards to code reviews and testing.
During the hackathon, the team developed three apps that were designed to improve app quality and streamline processes using AI. The following are the details of each app.
App #1: AI Buddy
The first app, named AI Buddy, allows users to ask ChatGPT to assist in the review, explanation or improvement of a Pull Request. The user interacts with the AI through Pull Request comments. The app responds to comments that are addressed to it and has access to the full context of the Pull Request to provide detailed and relevant input.
Through AI Buddy, the team discovered that ChatGPT is exceptionally helpful when asking questions about the content. However, typical-sized pull requests tend to be too big for the token limit of GPT 3.5, but this issue can be (partially) resolved with GPT 4.0. ChatGPT is also very good at explaining small chunks of code, and its API is easy to use.
AI generated app logo:
App #2: AI Chat Assistant
The second app, an extension of AI Buddy, is an AI chat assistant that helps users write helpful Pull Request titles and descriptions when creating a Pull Request.
The team found that ChatGPT can write short summaries, such as a title for a pull request, quite effectively. However, the longer summaries were sometimes off-topic and tended to halluzinate.
What helped a lot was providing additional context to the commit messages like the linked Jira issue descriptions and the code.
AI generated app logo:
App #3: Test Case Buddy for Jira
The third app, called Test Case Buddy for Jira, automates the creation of test cases for a specific Jira issue. It generates test cases in a table format, including the description of the test case, steps to perform the test, input data (if relevant), expected output, and a results column for users to fill in.
AI generated app logo:
Exploring the Limitations of AI
Through the development of these apps, we gained valuable insights into the potential of AI to transform the way we work and improve the quality of our apps. We also discovered some of the challenges and limitations of using AI, such as token limits and long response times, which we hope to overcome in future developments.
Conclusion
At Mibex Software, we believe that the potential of AI is limitless. We are committed to staying ahead of the curve by exploring new opportunities and pushing boundaries to drive innovation and growth.
We recognize there is always room for improvement, and we are determined to harness the full potential of this transformative technology to create solutions that benefit our customers.
We presented our learnings at a workshop of the GreaTest Quality Conference on June 1st, at Kraftwerk Zurich!