A system administrator may change what issue breakdown templates are available for selection and how breakdown is performed. This allows support for various different use cases, specific to your organization. To change the default behaviour, you may override the settings of the existing templates by modifying the following parameters:

• Enabled – if the breakdown template should be enabled
• Template Name – name of the breakdown template to be shown in the generate issues prompt
• Issue Prefix – optional prefix for newly generated issues. Use {PARENT_KEY} to replace with the parent issue key. Use {SEQUENTIAL_NUMBER} to replace with the sequential number.
• AI Query – instructions on how the breakdown should be performed, by specifying what should the issue description contain. Don’t specify the AI output formatting.

Breakdown is always done based on issue summary and issue description, through the AI query.

You may create additional breakdown templates.

Private AI Mode Integration Script

With Private AI mode enabled, you can bring your own AI model as an issue generation engine, used by the application. You will need to configure a simple Groovy integration script, that defines the communication protocol with the AI engine. Without configuring the script, the app will not function properly and will not be able to generate issues. By default, a sample script is available that serves as example – it can be modified or adapted. 

The script will be invoked every time when the Generate Issues form is submitted. 

Script invocation lifecycle is as follows:

1. Generate Issues form is submitted
2. Groovy script execution is triggered
3. Groovy script integrates with a Private AI model over HTTP or other means
4. Groovy script returns a single String result, which contains the generated issues

For proper execution the following best practises must be followed:

• The script must always return the text response from the AI engine as a String result.
• A Groovy parameter query is passed. It contains the AI query – it must be forwarded without change to the AI engine.
• In case an error occur, an Exception should be thrown. Example: throw new Exception(“An error has occured…”)
• Always ensure you put a timeout to all blocking operations, especially HTTP calls.

The following variables are available for use within the script:

• query – query to send to the AI model
• issueContext – the entire context of an issue, as string, to be sent as AI context. It includes parent Epic and issues with “relates to” links.
• language – selected language
• httpClient – used to easily send HTTP requests
• logger – may be used for logging. Use as: logger.info(“message”) | logger.debug(“message”) | logger.error(“message”). The resulting logs will be available in the standard atlassian-jira.log files – please ensure you enable the required level of logging for the com.deview_studios.atlassian.jira.dc.ai_issue_generator package from the Logging and profiling menu.

It’s recommended to enable DEBUG logging while developing the script to see if you’re receiving errors. Please see section Logging. 

If you need help developing the script, don’t hesitate to reach to a Deview Studios representative through our standard channels. No additional charges apply.

/* MODIFY THIS SCRIPT TO INTEGRATE YOUR AI MODEL. THIS IS JUST AN EXAMPLE. */
import com.atlassian.jira.util.json.*
import org.apache.http.entity.StringEntity

def payload = new JSONObject() // Prepare body as per API specification
    .put("model", "gpt-4o")
    .put("messages", new JSONArray()
    	.put(new JSONObject()
            .put("role", "system")
            .put("content", issueContext))
        .put(new JSONObject()
            .put("role", "user")
            .put("content", query)))
    .put("temperature", 1.0)
    .toString();

logger.debug("Preparing POST body: " + payload)

def post = httpClient.newPost("https://api.openai.com/v1/chat/completions") // Create a new request to specified URL
post.setEntity(new StringEntity(payload, java.nio.charset.StandardCharsets.UTF_8)); // Set HTTP payload
post.addHeader("Content-Type", "application/json"); // Set HTTP header
post.addHeader("Authorization", "Bearer REPLACE_ME"); // Set HTTP header

JSONObject json = new JSONObject(httpClient.sendRequest(post)) // Parse response as JSON
def result = json.getJSONArray("choices").getJSONObject(0).getJSONObject("message").getString("content") // Traverse JSON tree

return result // This script must always return the produced result

Usage

 

Limitations

AI Issue Breakdown Assistant for Jira has the following limitations:

• N/A