How to add lookup table data in Confluence to be used by Rovo

Question

Our organization finds Rovo very useful, we would like to give it more smarts about integral company data.

We have key data identifiers for our data and we would like to be able to find tickets related to either the identifiers or the data details using Rovo.

We exported the data and imported it into Confluence and Rovo was not able to interpret the data.

Rovo found the CSV file but could not parse it for data.

Any ideas on how we can improve this scenario.

Laura Campbell · Accepted Answer

Hi @Anne Gallant ,

I see on this page https://support.atlassian.com/rovo/docs/using-rovo-chat-for-data-analysis/ CSV files are supported for data analysis in Rovo chat, but it also says "Currently unstructured data is not supported. Some examples of unstructured data are Confluence pages, PDFs or Word documents, or Slack threads and emails." which makes me wonder if Rovo has as a hard time using a CSV file embedded on a Confluence page.

Have you tried adding the CSV file directly to a Rovo chat to see if it could parse the data?

If that doesn't work, if you have a paid Atlassian instance, you might want to raise a support request, or come to the Rovo Office Hours

Separate_Remove · Accepted Answer

Hello @Anne Gallant, fear not!I ran into something similar recently. Here are two approaches that *don’t* exactly solve the problem, but might get you close enough 🤣

Instead of attaching the CSV, paste the contents into a table in that same page. Then just explain in the Scenario what each column means, how the info should be used, and how the data sources connect. The Agent will easily do the rest - maybe, just ensure you cite both the page name and URL if the Agent has access to whole organizations or big spaces knowledge.

This is unfortunately only feasible if you can afford a bit of manual maintenance from time to time and / or your data doesn't change too often...

For more complex tasks I actually follow @Laura Campbell approach.

I can't cover the whole process but we basically simply drag freshly baked extractions in CSV into the chat - providing the Agent with a ton of info covering the device's uptime, user interactions, success rates etc - and have him do the ticket triage, categorization and preliminary "reality check".

We'll get Premium next year so for the moment it's basically a man-powered API,

The results you can get with just a tiny bit of preparation, a generous dose of sleep deprivation and a 600+ lines prompt are actually stunning - especially for a no-code Agent (it even made me consider freelancing again to jump on this train while it's still relatively new lol).

Be precise with the sheets and columns names (as far as I could see, the classic 'sheet_name:column_name' is enough) and split the process in numbered steps to avoid deviations from your workflow and streamline the process.