Where is data stored that is used by a PVA

(0) Share

Report

Posted on by jimirillo

Hi,

I have a few questions regarding data used by PVA.

We are considering building a Bot for internal use and giving it access to our internal SharePoint sites.

How does PVA consume and reference the SharePoint data? Is it copied to Data Verse? If so, is all the data copied in a sort of topic form or is it just indexed?

If copied, how long after the data is changed on SharePoint will the PVA reflect those changes?

Is there an ETA on when the generative answers capabilities will be available in environments created in regions other than the US?

Any answers to any of the questions would be appreciated.

James.

Categories:

General topics

I have the same question (0)

All responses (4)

Answers (1)

Pstork1 68,697 Most Valuable Professional on at

Like (0)

Report
Variables generated in a bot are stored in memory and deleted when the bot session ends.

Accessing data in SharePoint, or any other data source, will be done by calling a Power Automate flow. Once the data is processed it will be returned to the bot as a variable. But the bot itself doesn't have the ability to process or store data. The one exception to this would be the new generative answers feature that retrieves information from a site like SharePoint using AI. But that data is also just returned and displayed to the user by the bot its not stored permanently as far as I know.

If you use a flow to make changes to the data in SharePoint those changes should be available immediately. Other than system processing time there is no delay.

Not sure what the timetable is for the rollout of generative answers elsewhere.

Was this reply helpful? Yes No
jimirillo 7 on at

Like (0)

Report

Hi, thanks for the reply.

It is the generative answers I'm curious about.

If the SharePoint data is not stored or indexed, that would suggest that each time the PVA has to respond to a question who's answer lies in the SharePoint, it would need to trawl through all of the SharePoint pages, documents etc (and there will be a lot of them), apply language processing before it can provide a response. That doesn't sound like it would be very performant and would require a lot of processing.

Has anyone tried this and can comment on the performance?

Thanks

James

Was this reply helpful? Yes No
Verified answer

Pstork1 68,697 Most Valuable Professional on at

Like (1)

Report

Here's a quote from the documentation:

"Content analysis and retrieval happens on behalf of the signed-in user, meaning this feature requires user authentication to be configured for your bot using Azure AD under the Manual (for custom website) authentication option. The No authentication and Only for Teams authentication options can't retrieve information from Sharepoint or OneDrive for Business."

Since it is done using the security context of the user it has to be done at runtime. Elsewhere in the documentation it also specifies that it will leverage SharePoint and Bing Search to process the generative content. So everything I read says that it is always done when requested from the original URL or search engines authorized to crawl those URLs.

Was this reply helpful? Yes No
jimirillo 7 on at

Like (0)

Report

Thanks for your swift response Pstork1. That's what I needed.

Was this reply helpful? Yes No