Natural Search for Jira

• 

  Crafting JQL

• 

  Try generated JQL soon

Overview

This app generates Jira Query Language (JQL) queries from natural language instructions, enabling users to effectively utilize Jira. By teaching AI the rules of JQL and facilitating interactive query creation, users can quickly compose and execute queries. It simplifies the complexity of creating JQL, enhancing the value of using Jira.

Background

JQL is used throughout Jira, and the ability to create the right JQL is essential to maximize the value of Jira. However, creating JQL as intended is not easy and requires a significant amount of learning.In addition, as shown below, JQLs created using ChatGPT alone are often inaccurate.

Ex.1

• Requirements: Search for issues that have attachments
• This app: ✅attachments IS NOT EMPTY
• Chat GPT: ❌ issueFunction in attachmentsExist()

Ex.2

• Requirements: Find requests that require or required approval by jsmith
• This app: ✅approvals = approver("jsmith")
• Chat GPT: ❌ (issueFunction in linkedIssuesOf("assignee = jsmith and resolution is empty", "requires approval") OR issueFunction in linkedIssuesOf("assignee = jsmith and resolution is empty", "required approval"))

Ex.3

• Requirements: Find new issues created since the 14th of this month
• This app: ✅created > startOfMonth("+13d")
• Chat GPT: ❌ created >= startOfMonth("+14d")

How It Works

This app lists JQL rules in prompts, and the Language Model (LLM) interprets user requirements based on them. Challenges include compressing JQL rules and the fact that generating accurate JQL is difficult with just the rules specified in the documentation. To address these issues, innovative solutions have been applied, resolving the majority of problems.

What's next for Natural Search for Jira

We would like to be able to create more precise JQLs by taking into account the projects and customfields information about the user Jira instance.

Built With

• javascript
• openai