Data and deliveries
Where the data comes from
Section titled “Where the data comes from”When you ask the agent to show you a report, there are two different things it can show. The agent always tells you which one you are looking at.
| Live preview | Delivered file | |
|---|---|---|
| What it is | The export run now, by the agent | The file a scheduled run produced and delivered |
| When it is available | Always | Only for reports delivered as a download link, through the API, or as a webhook file, and only until the download link expires |
| Whose permissions | Yours: it shows only work items you can see in Jira | The permissions the schedule ran with: its owner’s, or the app’s if the schedule generates with add-on permissions |
| Size | A limited number of work items. The agent tells you when it is showing only part of the data | The whole file, read a page at a time |
Reports delivered by email or SFTP are not stored by Advanced Export, so for those the agent can show a live preview, but not the file that was sent. This is the same as the dashboard: see Your data on our servers.
Datasets
Section titled “Datasets”A live preview shows only the first work items, and the agent reads delivered files a page at a time. Neither is a good way to answer “how many” or “how long” over hundreds of work items: a language model that counts rows itself can miss some, and can give a different answer each time.
For questions like these the agent builds a dataset instead, and answers with SQL:
- What it holds: the work items of an export or a JQL query, with your Jira permissions, and by default their change history. Up to 20,000 work items and 500,000 history items. If a limit is reached, the agent tells you.
- How it answers: the agent writes a SQL query, the database calculates the result, and the agent explains it. Ask it to show the query, and you can check the logic or reuse it in a SQL export. The same query on the same dataset always gives the same numbers.
- Who can use it: only you. Nobody else, including other people’s agents, can query it.
- How long it lasts: 4 hours by default, at most 24 hours. Then it is deleted. You can also ask the agent to delete it sooner.
Building a dataset takes from seconds to a few minutes, depending on the number of work items. You can build one at a time, and up to 10 an hour.
Datasets are available for sites whose data is processed in the US region. A Jira admin can turn them off in App settings.
One-off deliveries
Section titled “One-off deliveries”An export has one schedule: the one the dashboard shows and edits. If you ask the agent to schedule an export that already has one, it offers to change that schedule instead. It cannot add a second one.
When the agent sends an export once (“send this to Jane now”), Advanced Export records it as a one-off delivery of that export. It never repeats, and it cannot be switched on as a schedule. It does not replace the export’s schedule in the dashboard, and its run is not shown in the export’s execution log. Ask the agent whether it succeeded, when it ran, and about any error. To send it again, ask the agent.
Data and privacy
Section titled “Data and privacy”The agent works on the same data as the rest of Advanced Export, processed in the region your admin chose. It does not add any new place where your data is stored: live previews are generated on request and not kept, and delivered files are the ones your schedules already store for download links. See Data Security and Privacy.
What you send to Rovo, and what the agent answers, is handled by Atlassian Rovo under Atlassian’s terms.