Tag: LowCode

Adding intelligence to Dataverse using Dataverse AI functions

Hi Folks,

While intelligence with the use of AI is being embedded into each and every part of the Microsoft ecosystem, it is good to know the features coming in the Power Platform space.

In this blog post, let’s see how we can use Dataverse AI Functions, their usage, advantages which can greatly ease summarizing, classifying, extracting, translating, assessing sentiment, or drafting a reply for common business scenarios.

To illustrate it better, I used a different AI Function for Canvas App, Model Driven App and Power Automate, hope you can follow the same for others as well.

What are Dataverse AI Functions?

Think of Dataverse AI Functions as prebuilt AI Functions which will add intelligence in your Apps and Flows without need to collect, build and train the models. They can be used in many places such as AI Builder, Power Automate, Power Apps, Low Code Plugins. Following are the AI Functions available…

  1. AIReply – Drafts a reply to the message you provide.
  2. AISentiment – Detects sentiment for the text you provide
  3. AISummarize – Summarizes the text you provide
  4. AIClassify – Classifies the text into one or more categories, you can use this from custom copilot
  5. AIExtract – Extracts specified entities like Names of people, phone numbers, places etc.
  6. AITranslate – Translate text from another language

Let’s see their usage in our favorite Canvas Apps first as illustration is easy and later in the post, I will mention how you can call the Dataverse AI Functions from Model Driven Apps, Power Automate so that you can get the real essence of it….

Utilizing ‘Dataverse AI Functions’ in a Canvas App

Create a new Canvas App and add ‘Environment’ Datasource as shown below.

All the ‘Dataverse AI functions’ can be accessed by ‘Environment‘ as shown below.

Let’s try the AIReply function in Canvas Apps

Add a textbox for storing the prompt or input string and a button control.

On the ‘OnSelect‘ event of the button, use the following formula to store the response in the AIResponse context variable, make sure you name the variables appropriately in your formula as per your naming defined in canvas apps.

UpdateContext({AIReplyResponse:Environment.AIReply({Text:AIInput.Text})})
view raw Dataverse AI Function hosted with ❤ by GitHub

Now create one more text variable to store the response and change the Default value to AIReplyResponse.PreparedResponse.

Try testing the app by providing inputs as below…

You should get a response from AIReply in the response field, you can try out other functions providing the necessary parameters required.

Utilizing ‘Dataverse AI Functions’ in a Power Automate

In Power Automate, all you can do to call Dataverse AI Functions is call the Unbound Action as below.

Passing the relevant input parameters is enough to get the output from these functions.

Let’s try AISentiment

Click on test, you should a response from Power Automate with the sentiment

Utilizing ‘Dataverse AI Functions’ in a Model Driven Apps

Do you want to utilize the similar capabilities of Dataverse AI Functions inside your custom code like in Plugins, Actions etc..

Let’s try AIClassify

var request = new OrganizationRequest("AIClassify")
{
["AllowMultipleCategories"] = false,
["Categories"] = titles,
["Text"] = classifyText
};
var result = service.Execute(request);
view raw DataverseFunctions.cs hosted with ❤ by GitHub

It was pretty much similar in AIBuilder as well…

Please do note that there are quotas to use these AI Functions at tenant level, else you might get similar error like below, while I didn’t get any information regarding this from Microsoft, so I am unsure about this as of writing this post, I will keep this updated if I get to know.

Using Dataverse AI functions needs a bit of Prompt Engineering knowledge, you were looking to learn more about Prompt engineering, then check it out here.

References:

https://learn.microsoft.com/en-us/power-platform/power-fx/reference/function-ai

Cheers,

PMDY

Dataverse Low Code Plugins – Insight

Prerequisite: Intall Dataverse Accelerator from AppSource from Microsoft Power CAT Team. Follow along once you have installed…see Dataverse Accelerator Installation Post if you want to know how..

Firstly, open the environment from https://make.powerapps.com where you would like to test this out, install the Dataverse Accelerator Solution as described above.

You should see there is an option to create Plugin from UI itself, no need to open Visual Studio and write your .Net Class library and register using Plugin Registration tool.

Please note that this is still an Experimental Feature and lot of improvements yet to be made..below is my quick test on basic CRUD Operation using this Plugin..

There were two types of Low Code Plugins

  1. Instant Plugins
  2. Automated Plugins

Let’s see how they work by creating an Instant Plugin…

Step 1: Click on New Plugin Button available

Step 2: Key in the display name and description, both are mandatory to be entered…we don’t have a way to customize here as of now…

Step 3:Then Click on Next and enter the Parameters in the Definition area…I just mentioned In and Out parameters for brevity, then added a simple expression to multiply the input variable supplied by 100…

Make sure you won’t get any errors

Step 4: Then click on Next and click on Save.

Step 5:Once Saved, you will get a new button for testing…

Step 6:Clicking on Test will get you to the below screen to enter your input parameter…

Step 7:Clicking on Run which gives the output in Web API and highlighted is the response received with a success.

Note:

For some of the Automated Plugins, got a failure message with no way to troubleshoot what was the error was…I have raised this to Microsoft, but they mentioned that these features are work in Progress…

Hope we get all the complex functionality there in Dynamics 365 CE Plugins be available within Dataverse Low Code Plugins with the help of Power Fx too in future….the App development is going to be lot more easier…with Microsoft Power Platform… and citizen developers can pitch in…

Reference: Dataverse Low Code Plugins Repo in GIT Hub

Thank you for reading…

Cheers,

PMDY

Power FX…get back your power with excel like formulas… A brief review…

Microsoft Power Fx is an open-source programming language that hopes to make coding as straightforward as building a spreadsheet. Wondered why Microsoft is bringing this excel backbone language to Power Platform range of Products & Services.

It is planned to expand Power Fx throughout the Power Platform over the next two years, including model-driven Power Apps, Microsoft Dataverse and Power Virtual Agents, Microsoft’s low-code chatbot platform. Developers will also be able to use it to directly edit apps in text editors like Visual Studio Code.

So Power Fx is the new name for the formula language for canvas apps. It’s a general-purpose, strong-typed, declarative, and functional programming language. Power Fx will be made available as open-source software. And source code can be found at this link..

When a formula is modified or a value is entered, the worksheet is immediately recalculated to reflect the changes. 

Let’s look more in-depth at all the things that Power Fx is doing for us and the freedom it has to optimize because the formula was declarative:

  • Asynchronous: All data operations in Power Fx are asynchronous. The maker doesn’t need to specify this, nor does the maker need to synchronize operations after the call is over. The maker doesn’t need to be aware of this concept at all, they don’t need to know what a promise or lambda function is.
  • Local and remote: Power Fx uses the same syntax and functions for data that’s local in-memory and remote in a database or service. The user need not think about this distinction. Power Fx automatically delegates what it can to the server, to process filters and sorts there more efficiently.
  • Relational data: Orders and Customers are two different tables, with a many-to-one relationship. The OData query requires an “$expand” with knowledge of the foreign key, similar to a Join in SQL. The formula has none of this; in fact, database keys are another concept the maker doesn’t need to know about. The maker can use simple dot notation to access the entire graph of relationships from a record.
  • Projection: When writing a query, many developers write select * from table, which brings back all the columns of data. Power Fx analyzes all the columns that are used through the entire app, even across formula dependencies. Projection is automatically optimized and, again, a maker doesn’t need to know what “projection” means.
  • Retrieve only what is needed: In this example, the LookUp function implies that only one record should be retrieved and that’s all that’s returned. If more records are requested by using the Filter function—for which thousands of records might qualify—only a single page of data is returned at a time, on the order of 100 records per page. The user must gesture through a gallery or data table to see more data, and it will automatically be brought in for them. The maker can reason about large sets of data without needing to think about limiting data requests to manageable chunks.
  • Runs only when needed: We defined a formula for the Text property of the label control. As the variable selected changes, the LookUp is automatically recalculated and the label is updated. The maker didn’t need to write an OnChange handler for Selection, and didn’t need to remember that this label is dependent upon it. This is declarative programming, as discussed earlier: the maker specified what they wanted to have in the label, not how or when it should be fetched. If this label isn’t visible because it’s on a screen that isn’t visible, or its Visible property is false, we can defer this calculation until the label is visible and effectively eliminate it if that rarely happens.
  • Excel syntax translation: Excel is used by many users, most of whom know that the ampersand (&) is used for string concatenation. JavaScript uses a plus sign (+), and other languages use a dot (.).
  • Display names and localization: First Name is used in the Power Fx formula while nwind_firstname is used in the JavaScript equivalent. In Microsoft Dataverse and SharePoint, there’s a display name for columns and tables in addition to a unique logical name. The display names are often much more user-friendly, as in this case, but they have another important quality in that they can be localized. If you have a multilingual team, each team member can see table and field names in their own language. In all use cases, Power Fx makes sure that the correct logical name is sent to the database automatically.

These no-code tools have been built to read and write Power Fx to ensure that there’s plenty of headroom for someone to take it further.

Now lets see what’s in there for developers here…

Professionals want to use professional tools to be most productive. Power Fx formulas can be stored in YAML source files, which are easy to edit with Visual Studio Code, Visual Studio, or any other text editor and enable Power Fx to be put under source control with GitHub, Azure DevOps, or any other source code control system. Take a look at below…

What a cool feature that Microsoft introduced all the excel functions and formulas to make low code apps. much easier ..if you want to further read regarding this…Microsoft Learn is the right place…this link can take you there…

Edit: 20/02/2022

Checkout the below image to see how power fx# reduces the length of code, a developer need to write…

That’s it for today…

Cheers,

PMDY