microsoft / coe-starter-kit

Other
751 stars 221 forks source link

[CoE Starter Kit - QUESTION] How do Identify Flows with AI Connectors #2362

Open binaryjam opened 2 years ago

binaryjam commented 2 years ago

What is your question?

I have some flows that are using AI Builder Actions, from the "AI Builder" Connector.

Well this seems to be nothing more than "Magic Numbers" from MS, the actual Connector listed and that makes the connection is the Dataverse Connector.

I need to be able to identify what is Actually DV Connectors and AI Builder Connectors. I've looked in the Flow Action Details, but this isnt what I thought, and is more of a log of all the actions over time, and even then it only lists entries a OpenApi Connection, no information on whether that was a read rows or a Get sentiment analysis.

How can we determine the user of DV versus AI Builder in the COE Toolkit, or even is there an existing admin connector, so that I can add a routine myself to identify the difference ? Thanks.

What solution are you experiencing the issue with?

Core

What solution version are you using?

November 2021

What app or flow are you having the issue with?

No response

AB#1643

manuelap-msft commented 2 years ago

At the moment, as - like you said - this uses the Dataverse connector, we don't have a way to identify what is AI Builder. But this is a great requirement, and we'll start investigating how we can surface that in the kit! Will keep you updated!

I also wanted to take a moment to really say a big THANK YOU for always taking the time to provide feedback on how to enhance the kit, it's making the kit better release by release!

binaryjam commented 2 years ago

I asked on twitter and the actions from get flow as admin contains a property that says the action is aibuilder.

I'm trying to use that myself, but it's gonna be tricky splitting AI and real DV apart.

On Tue, 5 Apr 2022, 20:06 Manuela Pichler, @.***> wrote:

At the moment, as - like you said - this uses the Dataverse connector, we don't have a way to identify what is AI Builder. But this is a great requirement, and we'll start investigating how we can surface that in the kit! Will keep you updated!

I also wanted to take a moment to really say a big THANK YOU for always taking the time to provide feedback on how to enhance the kit, it's making the kit better release by release!

— Reply to this email directly, view it on GitHub https://github.com/microsoft/coe-starter-kit/issues/2362#issuecomment-1089204943, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABYDEIEBBYU5C5DGKMFEOADVDSFLNANCNFSM5SPWUUBA . You are receiving this because you authored the thread.Message ID: @.***>

manuelap-msft commented 2 years ago

Oh, interesting! We'll explore that. I know AI models are also stored in the DV table of the environment (I think called aimodules) but from what I've seen it doesn't tell you where the model is used so can't correlate it from there.

Jenefer-Monroe commented 2 years ago

Hello we do track this already today in Flow Action Details table image

Can you take a look at that and see if it answers your question?

binaryjam commented 2 years ago

Hi Jenefer, this is intersting to know, I have a personal governance flow that checks for this and is reading the actions from PowerApps connector, so this would be easier, might revist that.

However the overall problem isnt one of the the creation of the COE Toolkit, but by the Product team.

To all our users and for use as admins, the AI Builder Connector appears as a Connector, we consider it as such as well, it is governed by its own credits also, nothing to do with Datverse, but for some reason known to them the product team have used this "Magic Number" concept on the actions (not even the connector) for dataverse.

Our reports and processes work via connectors, as Admin we want to know what apps are using AI Builder, not Dataverse and the Reports do not show this.

I know this is a tricky one, to actually do something here, the team needs to make a consideration, what's important here to our users, Should the COE Toolkit, make a "Mock AI Builder" Connector so that it can be reported , with the pain that comes with that of inspecting the actions and determining if 1 or both are used.

Whilst of course I could raise a ticket with the prod team, they will undoubtedly come back with, you can tell the difference via the actions, as you have, but that doesnt give us a reporting system that works as the users would expect.

What do we believe the user base wants in the long run ? Can COE Team influence the prod team and kick them , because they have Approvals as a connector which ustilises dataverse undeer the hood, why is AI builder different. But at least we can report on Approval vs Dataverse.

I would like to see the prod team fix this nonsense, so COE changes nothing, but if that does not happen , I would perfer a Mock Connector in COE for AI Builder so we can properly monitor its usage. Your choice in the end.

Jenefer-Monroe commented 2 years ago

Manuela has sent some feedback to the product team. Marking as product blocking.

azuindow commented 11 months ago

Hi there, would like to know if there are any updates regarding this matter from both the Production side and the CoE side.

Currently, our workaround involves using "Get Flow as Admin" to iterate through the environment and identify all the flows that utilize AI models. However, it would be more convenient if there was an original field or identifier for this purpose.

giramS09 commented 1 month ago

Hi @Jenefer-Monroe, would like to know is there any update on above issue i.e., if there is any field or identifier to differentiate between AI builder and Dataverse connector?

giramS09 commented 1 month ago

Hi @Jenefer-Monroe, could you please provide me with an update on above issue i.e., I am looking for if there is any field or any specific identifier (ID) to differentiate between AI builder and Dataverse connector?

Jenefer-Monroe commented 1 month ago

Sorry this is something that @Grant-Archibald-MS is driving as he is the one connected with the product team so I will need to defer to him.