Summary Suggester

App logo

Inspiration

In our job we very often faced situation when we started creating issue with detailed description, but at the end we had difficulties with coming up with simple, yet readable summary for this issue. That's where came up the idea that maybe we could create app to automate such things.

What it does

Our app works basically generates summary suggestions for Jira issue, based on provided informations, such as issue description, issue comments, parent issue description and parent issue summary. It can work both during issue creation (in create issue view, using UI modifications https://developer.atlassian.com/platform/forge/manifest-reference/modules/jira-ui-modifications/ ) and for already created issues (in issue view, using issue panel https://developer.atlassian.com/platform/forge/manifest-reference/modules/jira-issue-panel/ )

How we built it

We're using Atlassian Forge platform. Our app base on Forge's Custom UI - in our case, it's TypeScript-typed React app. We're using multiple Forge modules, such as admin page, issue panel and ui modifications.

Challenges we ran into

One of the biggest challenges for us was to prepare proper prompt for AI model, so it would always give issue summarization results, related to provided data, formatted every time in exactly same way.

Accomplishments that we're proud of

Aside of AI model itself, which is hosted on Azure OpenAI, we've been able to avoid any external infrastructure, which reduced both costs and architecture complexity.

What we learned

We significantly increased our knowledge both in developing apps on Atlassian Forge platform and in making use of AI models in our applications.

What's next for Summary suggester

We've been thinking about extending not only criterias considered on issue summary generation, but also extending list of fields that we'd like to generate suggestions for, like issue description.