ML experiment: Summarizing issue comments

Melissa Ushakov, Taylor McCaslin ·
Apr 13, 2023 · 3 min read

This blog post is part of an ongoing series about GitLab's journey to build and integrate AI/ML into our DevSecOps platform. The series starts here: What the ML is up with DevSecOps and AI?. Throughout the series, we'll feature blogs from our product, engineering, and UX teams to showcase how we're infusing AI/ML into GitLab.

GitLab issues are essential for team collaboration and serve as the source of truth for teams to align on the problem definition and scope of work for ongoing efforts. As teams collaborate on issues to refine them, the volume of comments grows. For issues with many comments, it can be challenging to understand the status of work at a glance. You may need to spend significant time reading comments to get an overview of the decisions made so far and to understand, for example, if there are any blockers.

With the recent advancements in AI and natural language processing, it's now possible for AI models to summarize text like that found in issue comments. We believe that this technology will help teams use their time more efficiently and help prevent losing track of important information within issue comments.

Large language models (LLMs) power generative AI solutions by using deep learning algorithms to analyze vast amounts of natural language data. These models are trained on massive datasets to develop an understanding of language patterns and context. Once trained, the models can generate new text that mimics human language.

In a rapid prototype, our own Alexandru Croitor, Senior Backend Engineer, and Nicolas Dunlar, Senior Frontend Engineer for our Plan stage, leverage generative AI LLMs to power comment summarization within GitLab's issues.

Prototype UX for comment summary

Above, you can see an example of triggering the summarization of issue comments. Watch the full demo below.

Our experiment takes an individual's natural language comments, inferences them against a generative AI LLM, and through novel prompt engineering (the task of guiding LLM output through instructions), creates a summary of long comment threads. Part of our engineering exploration is examining how to chunk extremely long comment threads into parsable bits an LLM can succinctly and accurately summarize.

Iterating on AI/ML features

While just an experiment today, we are iterating on how to effectively bring features like this to our customers. We're starting with summarization of issue comments, and are working to optimize prompts to provide more meaningful summaries. We are also investigating bringing this functionality to other objects like epics and merge requests.

This experiment is just the start of many ways we’re looking to infuse GitLab with AI/ML capabilities to help GitLab users become more efficient and effective at their jobs. We are looking across the software development lifecycle for painful and time-consuming tasks that are ideal for AI Assisted features. We’ll continue to share these demos throughout this blog series.

Interested in using these AI-generated features? Join our waitlist and share your ideas.

Continue reading our ongoing series, "AI/ML in DevSecOps".

Disclaimer: This blog contains information related to upcoming products, features, and functionality. It is important to note that the information in this blog post is for informational purposes only. Please do not rely on this information for purchasing or planning purposes. As with all projects, the items mentioned in this blog and linked pages are subject to change or delay. The development, release, and timing of any products, features, or functionality remain at the sole discretion of GitLab.

“Are your issue comment threads long? Wish there was a TLDR;? Learn how @gitlab is experimenting with ML-powered product features that help summarize issue comments.” – Melissa Ushakov, Taylor McCaslin

Click to tweet

Edit this page View source