Skip to content

45-Multi-Tenancy-Features-GenAI.md

Latest commit

 

History

History
271 lines (197 loc) · 16 KB

45-Multi-Tenancy-Features-GenAI.md

File metadata and controls

271 lines (197 loc) · 16 KB

Add Capabilities for Generative Artificial Intelligence (Gen AI)

Put yourself in the shoes of a poetry slam manager who uses a poetry slam management application to manage the events. You want to easily create events with creative titles and descriptions. For the title and description proposal, the Partner Reference Application uses generative artificial intelligence (generative AI or genAI) based on large language models (LLM). The SAP AI Core service of the SAP BTP and the SAP Cloud SDK for AI offer easy consumption of generative AI features. For information about getting started with the JavaScript SDK, see the Getting Started documentation. You can find information on security, data protection, and privacy aspects of the SAP AI Core service in the Security chapter of the service documentation.

AI Ethics

SAP has introduced a certification program for partner applications developed on SAP Business Technology Platform (BTP) using the generative AI hub. This program includes checks for Responsible AI compliance. It enables partners to offer trusted, compliant, and enterprise-ready applications powered by AI services, leveraging SAP’s expertise in business data insights. Additionally, the SAP Global AI Ethics Policy and the SAP AI Ethics Handbook provide guidance on applying the SAP AI Policies.

Furthermore, the SAP Generative AI Hub offers capabilities like Data Masking to support data protection and privacy implementation.

Bill of Materials

Entitlements

In addition to the entitlements listed for the multitenancy version, the list shows the entitlements that are required in the different subaccounts to add generative artificial intelligence.

Subaccount Entitlement Name Service Plan Type Quantity
Provider
SAP AI Core extended Service 1

Guide How to Enhance the Application Step by Step

To explore this feature with the Poetry Slam Manager, you have two options:

  1. Clone the repository of the Partner Reference Application. Check out the main-multi-tenant branch and enhance the application step by step.

  2. Alternatively, check out the main-multi-tenant-features branch where the feature is already included.

The following describes how to enhance the main-multi-tenant branch (option 1).

Application Enablement

  1. Enhance the poetry slam service.

    1. Enhance the poetry slam service definition.

      1. Add a new action createWithAI to the PoetrySlams entity. This action creates a new poetry slam event and uses generative AI to propose a title and description.

        @(cds.odata.bindingparameter.collection)
@(UI.IsAIOperation: true) // Add the AI Icon to the action button
action createWithAI(
    @(
        title: '{i18n>languageInput}',
        mandatory: true,
        Common:{
            ValueListWithFixedValues: false,
            ValueList               : {
                $Type         : 'Common.ValueListType',
                CollectionPath: 'Language',
                Parameters    : [
                    {
                        $Type            : 'Common.ValueListParameterInOut',
                        ValueListProperty: 'name',
                        LocalDataProperty: language,
                    },
                    {
                        $Type            : 'Common.ValueListParameterDisplayOnly',
                        ValueListProperty: 'code'
                    },
                    {
                        $Type            : 'Common.ValueListParameterDisplayOnly',
                        ValueListProperty: 'descr'
                    }
                ]
            },
        }
    )
    language : String(50),
    @(
        title: '{i18n>tagsInput}',
        UI.Placeholder: '{i18n>placeholder}',
        mandatory: true
    )
    tags : String(100),
    @(
        title: '{i18n>rhymeInput}',
        UI.ParameterDefaultValue: true
    )
    rhyme : Boolean, ) returns PoetrySlams;

      2. Add the Language entity from the module sap.common.Languages.

        //Languages
entity Language    as projection on sap.common.Languages;

    2. Copy the srv/lib/genAI.js file with the genAI class to your project. The genAI class uses SAP Cloud SDK for AI with the orchestration and prompt-registry package to leverage the generative AI hub features. Additionally, it defines which large language model is used for the proposal generation and defines the system prompt.

      Note: The implementation uses the Azure content filter for restricting content that is passed to and received from a generative AI model. Only safe content for input and output of different categories (such as hate) are allowed.

      Note: The implementation uses the masking module of the orchestration client to mask sensitive information in the prompt. It adds the masking provider SAP Data Privacy Integration, which anonymizes or pseudonymizes sensitive information depending on the replacement strategy.

    3. Extend the service implementation file srv/poetryslam/poetrySlamServicePoetrySlamsImplementation.js with the implementation of the action.

      1. Import the genAI class.

        const GenAI = require('../lib/genAI');

      2. Add the action implementation.

        // Entity action: Create a poetry slam with generative artificial intelligence
srv.on('createWithAI', async (req) => {
    const genAI = await GenAI.init();

    const response = await genAI.callOrchestrationChatCompletion(
      req.data.tags,
      req.data.language,
      req.data.rhyme,
      req
    );

    // In case the orchestration call could not be started, no draft will be created
    if (!response) return null;

    const poetrySlamDraft = await GenAI.createPoetrySlamWithAI(
      response,
      req,
      srv,
      db
    );
    return poetrySlamDraft;
});

    4. Add UI texts for the action parameter dialog into the file srv/i18n/i18n.properties.

      # -------------------------------------------------------------------------------------
# AI 

languageInput           = Select the language in which you want the title and description to be generated
tagsInput               = Add some tags to describe the generated title and description
rhymeInput              = Generate the description in rhymes
placeholder             = For example: creative, funny

      Note: In the reference example, the file srv/i18n/i18n_de.properties with the German texts is available, too. You can adopt them accordingly.

    5. Add the message texts for the action error handling into the file srv/i18n/messages.properties.

      ACTION_AI_NO_ACCESS                                     = Access to SAP AI Core service isn’t possible. Please reach out to your application provider.
ACTION_AI_MISSING_PARAMETERS                            = Please enter a language and tags. 
ACTION_AI_INVALID_PARAMETERS                            = Invalid parameters. Please check your input.
ACTION_AI_ORCHESTRATION_ERROR                           = An error occurred during AI orchestration. Check the logs for more details.
ACTION_AI_FILTER_VIOLATION                              = This content doesn't meet our safety guidelines. Please revise your input and try again.

      Note: In the reference example, the file srv/i18n/messages_de.properties with the German texts is available, too. You can adopt them accordingly.

  2. Enhance the SAP Fiori elements UI of the Poetry Slams application.

    1. Enhance the section LineItem of the file app/poetryslams/annotations to add a button that triggers the action to create the poetry slam with generative AI.

      {
    $Type : 'UI.DataFieldForAction',
    Action: 'PoetrySlamService.createWithAI',
    Label : '{i18n>createWithAI}'
},

    2. Add the UI text for the button into the file /app/poetryslams/i18n/i18n.properties.

      createWithAI            = Create with Slamtastic AI

      Note: In the reference example, the file /app/poetryslams/i18n/i18n_de.properties with the German texts is available, too. You can adopt them accordingly.

  3. Add the required npm modules as dependencies to the package.json of your project. Refer to the file package.json of the sample application.

    1. Open a terminal.

    2. Run the command npm add @sap-ai-sdk/orchestration. This package incorporates generative AI orchestration capabilities into your AI activities in SAP AI Core. It is part of the SAP Cloud SDK for AI, which is the official Software Development Kit (SDK) for SAP AI Core, the generative AI hub, and orchestration workflow.

    3. Run the command npm add @sap-ai-sdk/prompt-registry. This package incorporates generative AI prompt registry capabilities into your AI activities in SAP AI Core. It is also part of the SAP Cloud SDK for AI.

SAP BTP Configuration and Deployment

  1. Open the SAP BTP cockpit of the provider subaccount and add the required entitlements:

    • SAP AI Core with the extended plan to access the SAP AI Core service.

  2. Add the SAP AI Core service as a resource in the file mta.yaml. Besides this, the resource is required as dependency in the service module and the mtx module.

    modules:
    - name: poetry-slams-srv
      requires:
        - name: poetry-slams-aicore

    - name: poetry-slams-mtx
      requires:
        - name: poetry-slams-aicore

resources:
    # AI Core Service
    - name: poetry-slams-aicore
        type: org.cloudfoundry.managed-service
        parameters:
        service: aicore
        service-plan: extended

  3. Run the command npm install in your project root folder to install the required npm modules.

  4. Build and deploy the application.

    Note: For detailed instructions on how to deploy, refer to Deploy Your SAP BTP Multi-Tenant Application.

Unit Tests

Unit tests are available to test the artificial intelligence feature:

  1. Enhance the file test/srv/poetryslam/poetrySlamServicePoetrySlams.test.js with a test to check the action.

    it('should reject createWithAI action without running SAP BTP AI Core service', async () => {
    await expect(
        ACTION(`/odata/v4/poetryslamservice/PoetrySlams`, `createWithAI`, {
            language: 'English',
            tags: 'funny',
            rhyme: false
        })
    ).to.rejectedWith();
});

  2. Copy the file test/srv/lib/genAI.test.js to your project. This file tests the genAI class.

  3. To run the automated SAP Cloud Application Programming Model tests:

    1. Enter the following command in a terminal in SAP Business Application Studio to install all required node modules.

      npm install

    2. Enter the following command to run all tests. The result is shown afterwards.

      npm run test

A Guided Tour to Explore the Generative Artificial Intelligence Feature

Now it is time to take you on a guided tour through the generative artificial Intelligence feature of the Poetry Slam Manager:

  1. Open the SAP BTP cockpit of the customer subaccount.

  2. Open the Poetry Slams application.

  3. Choose Create with Slamtastic AI to create a new poetry slam with a proposed title and description from generative artificial intelligence.

  4. Select a language.

  5. Add one or more tags that describe the content and style of the proposed title and description. Examples are funny and creative.

  6. The parameter rhyme is defaulted. Stay with the default to get a text that is a rhyme.

Note: The object page with the newly created poetry slam is opened. The title and description are proposed accordingly.

Metering genAI Usage

The use of models in the generative AI hub of the SAP AI Core service is metered using genAI tokens. For more information, refer to the documentation on SAP Help Portal.

In case you want to estimate how many input and output tokens are required for the use case of the Partner Reference Application, the method callOrchestrationChatCompletion of class genAI shows how to determine the used tokens per genAI request. On this basis, you can make an assumption for your use case.