[[{“value”:”

Enhance SAP Joule for Consultants with Custom Knowledge Grounding

Table of Contents

• 1. Introduction
  • 1.1 Overview
  • 1.2 Recommendations
• 2. Data Repositories
  • 2.1 Data Repositories – Overview
  • 2.2 Object Store
  • 2.3 Microsoft SharePoint
  • 2.4 SAP Build Work Zone
• 3. AI Core Document Grounding
  • 3.1 AI Core Document Grounding – Overview
  • 3.2 AI Core instance creation
  • 3.3 AI Core Document Grounding with Bruno
  • 3.4 AI Core Document Grounding with AI Launchpad
  • 3.5 Support
• 4. Connection to SAP Joule for Consultants
  • 4.1 Destination Service setup
  • 4.2 Admin Console configuration
• 5. Testing
  • 5.1 Testing Custom Knowledge Grounding

1. Introduction

1.1 Overview

1.1 a) What Custom Knowledge Grounding Does

Custom Knowledge Grounding extends SAP Joule for Consultants with your organization’s own documents. The documents are uploaded to a data repository, indexed by SAP AI Core Document Grounding, and retrieved at runtime to ground SAP Joule for Consultants’ answers. Users get responses that combine the SAP-curated content SAP Joule for Consultants ships with and their firm’s own methodologies, policies, and delivery standards.

1.1 b) Prerequisites

• SAP Joule for Consultants — an existing instance. See Onboarding Guide for SAP Joule for Consultants | SAP Help Portal.
• SAP AI Core with the extended plan — required by Document Grounding. See SAP AI Core — SAP Discovery Center.
• Object Store (recommended for the initial setup) — the simplest data repository to start with, and the example used throughout this guide. See Object Store — SAP Discovery Center.

1.1 c) Pricing

Custom Knowledge Grounding relies on paid SAP services that are billed separately from SAP Joule for Consultants:

• SAP AI Core (extended plan) is mandatory. Charges accrue per document indexed and per token retrieved — see the Generative AI Hub in SAP AI Core Calculator and Pricing.
• Object Store (standard plan) if you take the recommended setup path. See Pricing.
• SAP AI Launchpad (standard plan) is optional — needed only if you prefer the UI-driven setup over the REST API path. See Pricing.

Review the pricing of each service before subscribing.

1.1 d) What This Guide Covers

The remaining sections walk through the end-to-end setup:

• 2. Data Repositories — pick a supported data repository (e.g., Object Store, Microsoft SharePoint, or SAP Build Work Zone) and upload your documents to it.
• 3. AI Core Document Grounding — provision SAP AI Core with the extended plan, then create a pipeline that ingests, chunks, embeds, and indexes the documents.
• 4. Connection to SAP Joule for Consultants — configure a Destination Service instance, then enable Custom Knowledge Grounding in the SAP Joule for Consultants Console and connect it to the pipeline.
• 5. Testing — verify the setup by asking SAP Joule for Consultants a question whose answer is only available in your uploaded documents.

Note: Depending on what is already in place — an existing AI Core Document Grounding pipeline, ready-to-use data repository credentials, or neither — you may be able to skip directly to a later section. The next section, 1.2 Recommendations, opens with a decision tree for choosing the right entry point and then covers data repository choice and subaccount layout.

1.2 Recommendations

1.2 a) Where to Start

Not every reader has to work through every section of this guide. Use the decision tree below to pick the entry point that matches what you already have in place:

• An AI Core Document Grounding pipeline is already provisioned (resource group, generic secret, and a pipeline reporting FINISHED against the data repository you intend to use) → skip ahead to the 4. Connection to SAP Joule for Consultants section, starting with 4.1 Destination Service setup. Configure the Destination Service against the existing pipeline and enable Custom Knowledge Grounding in the SAP Joule for Consultants Console.
• The data repository credentials are already available (e.g., a SharePoint site with the Entra app registered, an S3 bucket, or a Work Zone tenant with Document Grounding enabled) and you have run an AI Core Document Grounding setup before → skip the 2. Data Repositories section and start with the 3. AI Core Document Grounding section.
• Neither of the above applies → start at the 2. Data Repositories section with 2.2 Object Store, as described below. It is the fastest path to a working setup and the one the rest of this guide uses as the running example.

1.2 b) Start with Object Store

Custom Knowledge Grounding can read from any of the supported data repositories. For the initial setup, Object Store is the recommended choice:

• Fast activation — the Object Store instance is provisioned by a booster and ready in minutes. SharePoint, customer-managed S3, and Work Zone setups often take longer than anticipated because of onboarding complications, approval, and security processes.
• Very low cost — Object Store charges are small. Even when run for a few months until the target data repository is ready, the extra cost is negligible — see Pricing.
• Reversible — once the target data repository is online and indexed, point Custom Knowledge Grounding at the new pipeline and decommission the Object Store instance.

The key advantage is that Custom Knowledge Grounding can be live in SAP Joule for Consultants while the longer-running setup of your target data repository runs in parallel — instead of blocking the entire rollout on it.

Note: The Object Store instance must be provisioned in an AWS-based subaccount, since AI Core Document Grounding consumes it through the AWS S3 API.

Note: Answers themselves are generated the same way regardless of the data repository. Source citations are clickable for SharePoint and Work Zone — they link back to the original file. For Object Store and the other repositories, citations show the document name and page reference but do not resolve to the file.

1.2 c) Subaccount Layout and Existing Instances

The four moving parts — Object Store (or any other data repository), SAP AI Core, Destination Service, and SAP Joule for Consultants — communicate through service-key credentials, which allows for very flexible placement: they can all sit in the same subaccount, in different subaccounts, or even in different global accounts.

• Reuse existing instances where available. If your organization already runs SAP AI Core (with the extended plan) or SAP AI Launchpad in another subaccount, you can reuse them instead of provisioning new instances.
• The Destination Service can be deployed in the SAP Joule for Consultants subaccount or in the AI Core subaccount. In general, it is recommended to place it in the same subaccount as the consuming application—in this case, the SAP Joule for Consultants subaccount.
• The Object Store must live in an AWS-based subaccount (see above). The SAP Joule for Consultants subaccount is a convenient option, since it runs on AWS data centers.

2. Data Repositories

2.1 Data Repositories – Overview

2.1 a) What a Data Repository Provides

AI Core Document Grounding does not host the source documents itself — it reads them from an external data repository and serves the resulting index to SAP Joule for Consultants. The data repository is the system of record where you upload and maintain the documents you want to ground SAP Joule for Consultants’ answers on.

The starting point for the AI Core Document Grounding documentation is Grounding | SAP Help Portal.

2.1 b) Supported Data Repositories

AI Core Document Grounding can read content from the following data repositories:

• Microsoft SharePoint
• AWS S3
• SFTP
• SAP Build Work Zone
• SAP Document Management service
• ServiceNow
• Google Drive

Setup steps for Google Drive, Microsoft SharePoint, SAP Build Work Zone, and ServiceNow (SNOW) are described on Set Up Data Repository Integration | SAP Help Portal.

Note: That help page is written for Document Grounding with SAP Joule for Business. Only the data repository setup instructions apply here — the remaining content on the page is specific to SAP Joule for Business and is not relevant for SAP Joule for Consultants.

The next subsections walk through the setup of three data repositories in more detail:

• 2.2 Object Store
• 2.3 Microsoft SharePoint
• 2.4 SAP Build Work Zone

The three subsections are independent — pick the repository (or repositories) you want to use and skip the others. For any data repository not covered here, refer to the SAP Help pages linked above.

2.2 Object Store

2.2 a) Provision the Object Store Service

The Object Store provides S3-compatible storage where documents are uploaded for AI Core Document Grounding to index. Begin by verifying the entitlement at the global account and adding the service plan to your subaccount.

Note: Create the Object Store in an AWS-based subaccount; the steps below assume an AWS backend.

• At the global account level, open Entitlements → Service Assignments.
• Search for Object Store and confirm the standard plan is assigned.

• In your subaccount, open Entitlements and click Edit.
• Click Add Service Plans, select Object Store, check the standard plan, and click Add 1 Service Plan.
• Click Save to apply the changes.

2.2 b) Set Up the Cloud Foundry Environment

The Object Store instance runs in the Cloud Foundry runtime, so the subaccount must have Cloud Foundry enabled and a space available to host the instance.

• In your subaccount, open Overview and click Enable Cloud Foundry.
• Keep the Cloud Foundry Runtime environment and standard plan, adjust the Instance Name and Org Name if needed, then click Create.

• Once Cloud Foundry is enabled, click Create Space on the Cloud Foundry Environment tab.
• Enter a space name (for example, j4c), assign yourself the Space Developer and Space Manager roles, and click Create.

2.2 c) Create the Object Store Instance

• In your subaccount, open Instances and Subscriptions and click Create.
• Fill in the required fields:
  • Service: Object Store
  • Plan: standard
  • Runtime Environment: Cloud Foundry
  • Space: the space you created (e.g., j4c)
  • Instance Name: e.g., j4c-object-store
• Click Create to provision the instance.

• Click on the instance row to open its details panel on the right.
• Under Service Keys, click Create.
• Enter a name (e.g., j4c-object-store-sk) and click Create.

• Once the key’s Status shows Created, click the … menu on the service key row and choose Download to save the credentials as a JSON file. You can also choose View to inspect them in the browser and copy individual values.

Note: The service key contains the access_key_id, secret_access_key, region, and bucket values needed to upload documents — you’ll use them in the next step.

2.2 d) Upload Documents with the AWS CLI

With the service key credentials, you can use the AWS CLI to upload documents to the Object Store bucket. AI Core Document Grounding will then index these documents for retrieval.

• Install the AWS CLI if it isn’t already on your machine.

Note: If you prefer a graphical client, Cyberduck supports S3 with the same service key credentials.

• Download the sample file Test_document–Heliotrope_Fabricator.pdf. You can append --Object-Store to the filename (e.g., Test_document--Heliotrope_Fabricator--Object-Store.pdf) to mark which storage backend the copy belongs to.
• Run aws configure and paste in the credentials when prompted — the default output format can be left blank.
• Upload the document to the bucket and list its contents to verify:

aws s3 cp <file> s3://<bucket>
aws s3 ls s3://<bucket>

Note: The bucket name in the service key is prefixed with hcp- followed by a unique identifier.

2.3 Microsoft SharePoint

Disclaimer: The steps below are a reference. Coordinate with your SharePoint administrator and validate against the documentation below.

• Set Up Data Repository Integration | SAP Help Portal — the official SAP guidance for this setup. It is written for SAP Joule for Business, but the data repository setup steps apply here as well.
• Understanding Resource Specific Consent for Microsoft Graph and SharePoint Online | Microsoft Learn — Microsoft’s reference on the Resource Specific Consent permission model that the steps below rely on.

2.3 a) Register the Microsoft Entra Application

AI Core Document Grounding connects to a SharePoint site through a Microsoft Entra application using the OAuth2ClientCredentials flow. Start by registering the application in Microsoft Entra.

• Sign in to entra.microsoft.com with an account that can register applications.
• Open App registrations and click New registration.
• Enter a name (e.g., j4c-sharepoint-grounding), keep the default settings, and click Register.
• From the Overview tab, copy the Application (client) ID and the Directory (tenant) ID — you’ll need both for the 3. AI Core Document Grounding section.

2.3 b) Grant the Sites.Selected Permission

The application needs the Sites.Selected Microsoft Graph permission. On its own, this permission grants access to zero SharePoint sites — it only makes the application eligible to be granted access to specific sites in a later step. This is intentional and enforces least-privilege.

• In the registered application, open API permissions and click Add a permission.
• Select Microsoft Graph → Application permissions.
• Search for Sites.Selected, check it, and click Add permissions.

Note: Use the Sites.Selected permission listed under Microsoft Graph — not the one with the same name listed under SharePoint. They are different resources, and AI Core Document Grounding consumes Microsoft Graph.

• Click Grant admin consent for [Your Org]. A tenant Global Administrator or Privileged Role Administrator must perform this step.
• Verify that the Status column for Sites.Selected shows a green check mark.

2.3 c) Create a Client Secret

The OAuth2ClientCredentials flow authenticates with a client secret instead of a user sign-in.

• In the registered application, open Certificates & secrets and click New client secret.
• Enter a description, pick an expiry, and click Add.
• Copy the secret Value immediately — it is only displayed once. Store it together with the tenant ID and application ID.

2.3 d) Link the Application to the SharePoint Site

After admin consent, the application still cannot read any site. You now have to bind it explicitly to the SharePoint site that hosts your grounding documents. The easiest path is the PnP.PowerShell module.

Note: The account running the grant command must have the Site Collection Administrator role on the target SharePoint site (a tenant Global Administrator works as well). Owners of the Microsoft 365 Group or Teams team that backs a SharePoint site automatically have this role. If you don’t have it, ask a SharePoint administrator to grant it temporarily. Sites.Selected on its own is not enough to perform the grant.

• Install PowerShell 7.2 or later — see Install PowerShell on Windows, Linux, and macOS | Microsoft Learn.
• Install the PnP.PowerShell module — see Installing PnP PowerShell | PnP PowerShell.
• On first use in the tenant, register PnP’s own Entra application by running Register-PnPEntraIDApp and following the prompts. The resulting Client ID is the helper-app ID used to sign in with Connect-PnPOnline — it is not the ID of the application you registered in the previous sections.

Note: Register-PnPEntraIDApp is a one-time setup per tenant. Skip it if PnP.PowerShell has already been initialized.

• Sign in to the target SharePoint site and grant the Sites.Selected application read access:

  Connect-PnPOnline -Url "<site-url>" -Interactive -ClientId "<pnp-helper-app-client-id>"
  Grant-PnPAzureADAppSitePermission `
      -AppId "<sites-selected-app-client-id>" `
      -DisplayName "<display-name>" `
      -Site "<site-url>" `
      -Permissions Read

  Replace <site-url> with the full SharePoint site URL (e.g., https://contoso.sharepoint.com/sites/KnowledgeBase), <pnp-helper-app-client-id> with the helper-app client ID from Register-PnPEntraIDApp, and <sites-selected-app-client-id> with the Application (client) ID of the application you registered for AI Core Document Grounding.

• A successful response prints the permission ID, the granted role (read), and the application identity.

Note: Read is sufficient for AI Core Document Grounding. The same grant can also be issued through a Microsoft Graph POST /sites/{site-id}/permissions call — see Understanding Resource Specific Consent for Microsoft Graph and SharePoint Online | Microsoft Learn for the REST equivalent.

2.3 e) Upload Documents to the SharePoint Site

• Open the SharePoint site in a browser and upload your documents to a document library.
• Download the sample file Test_document–Heliotrope_Fabricator.pdf. You can append --SharePoint to the filename (e.g., Test_document--Heliotrope_Fabricator--SharePoint.pdf) to mark which storage backend the copy belongs to.

The Bruno and AI Launchpad subsections in the 3. AI Core Document Grounding section consume the following SharePoint credentials, which you have now collected:

• MS_Entra_Tenant_ID — Directory (tenant) ID from the Overview tab.
• MS_Entra_Application_ID — Application (client) ID from the Overview tab.
• MS_Entra_Application_Client_Secret — the secret value copied from Certificates & secrets.
• SharePoint_URL — the full URL of the SharePoint site you granted permission on.

2.3 f) Troubleshooting

The Troubleshooting Document Grounding Pipelines | SAP Community blog walks setup failures down in three steps:

• Step 1 — Validate access to the Document Grounding API.
• Step 2 — Validate SharePoint credentials.
• Step 3 — Validate pipeline configuration.

Especially noteworthy are the checks for step 3:

• SharePoint site identifier. Use the identifier from the site URL (e.g., HR%20Team for https://contoso.sharepoint.com/sites/HR%20Team), not the human-readable site title. Mind the URL encoding.
• Include paths. Match the URL path of the folder you want to ground on, relative to the document library — /teched, not /Shared%20Documents/teched.
• Pipeline created but no documents. Almost always means the include path is too narrow. Test with no include path first; once replication works, narrow it down.

2.4 SAP Build Work Zone

2.4 a) Provision SAP Build Work Zone, Advanced Edition

Document Grounding is available in SAP Build Work Zone, advanced edition across all plans. Provision the tenant if you don’t already have one.

• Follow the Onboarding to SAP Build Work Zone, advanced edition | SAP Help Portal guide.
• The community post SAP Build Work Zone: advanced edition — step-by-step configuration | SAP Community covers the same ground with screenshots and may be easier to follow on a first pass.

2.4 b) Enable Document Grounding on the Work Zone Tenant

Once the Work Zone tenant is up, enable Document Grounding so AI Core can ingest its content, and collect the OAuth credentials AI Core will use to read it.

• Follow Integration With Document Grounding | SAP Help Portal.
• The community blog Joule — Document Grounding with SAP Build Work Zone, advanced edition | SAP Community walks through the same enablement end-to-end.

Note: Stop after Step 3 of the community blog. Step 4: Use an Existing Joule Subaccount or Create a New One to Set Up Document Grounding is replaced by the AI Core Document Grounding setup in the 3. AI Core Document Grounding section of this guide — the Destination Service and SAP Joule for Consultants Console steps later on perform the equivalent integration for SAP Joule for Consultants.

2.4 c) Upload Documents to Work Zone

• In the Work Zone workspace configured for Document Grounding, upload your documents to the workspace folder.
• Download the sample file Test_document–Heliotrope_Fabricator.pdf. You can append --Work-Zone to the filename (e.g., Test_document--Heliotrope_Fabricator--Work-Zone.pdf) to mark which storage backend the copy belongs to.

The Bruno and AI Launchpad subsections in the 3. AI Core Document Grounding section consume the following Work Zone credentials, which you collected during the Document Grounding enablement above:

• DWS_URL — the URL of the Work Zone tenant.
• OAuthClient_Key — the OAuth client key.
• OAuthClient_Secret — the OAuth client secret.

2.4 d) Support

Issues specific to the Work Zone side of this setup — Document Grounding capability, OAuth configuration, content not appearing in the workspace folder — belong to support component EP-WZ-STA. Issues with the AI Core pipeline that reads from Work Zone (indexing, embeddings, pipeline status) belong under CA-ML-RAGE. The 3.5 Support subsection later in the next section has the full breakdown.

3. AI Core Document Grounding

3.1 AI Core Document Grounding – Overview

3.1 a) What AI Core Document Grounding Does

SAP AI Core Document Grounding reads documents from a data repository, chunks them, generates embeddings, and stores them in a vector index. SAP Joule for Consultants queries that index at runtime to ground its answers in your organization’s own content.

Note: For a conceptual introduction to what happens inside that pipeline — vector embeddings, the SAP HANA Cloud vector engine, and how data ingestion, orchestration, and retrieval fit together in the Generative AI Hub — see the Analyzing Document Grounding in Generative AI Hub | SAP Learning lesson. It is theoretical background, not setup instructions, and complements the hands-on subsections that follow.

3.1 b) What the Rest of This Section Covers

The next subsections walk through the setup of SAP AI Core Document Grounding. The instructions in both the Bruno and AI Launchpad subsections use the Object Store as the example data repository:

• 3.2 AI Core instance creation — provision SAP AI Core with the extended plan via the AI Core booster, and download the service key.
• 3.3 AI Core Document Grounding with Bruno — create the resource group, generic secret, and pipeline through the REST API using an importable Bruno collection.
• 3.4 AI Core Document Grounding with AI Launchpad — an alternative path that performs the same setup through a UI instead of the REST API.
• 3.5 Support — where to open tickets when something goes wrong.

Note: The Bruno and AI Launchpad subsections are two interchangeable paths to the same outcome. Bruno is a free, open-source API client; SAP AI Launchpad is an alternative UI-driven route that requires a paid subscription.

3.1 c) Pricing

SAP AI Core offers a free plan, but Document Grounding requires the extended plan, which is a paid subscription. Charges accrue per document indexed and per token retrieved, and are billed separately from SAP Joule for Consultants. Review the pricing before subscribing:

• SAP AI Core — SAP Discovery Center
• Generative AI Hub in SAP AI Core Calculator

If you take the AI Launchpad path, note that SAP AI Launchpad with the standard plan is also a paid subscription: SAP AI Launchpad — SAP Discovery Center.

3.1 d) Supported Data Repositories

AI Core Document Grounding can index content from the following data repositories:

• Microsoft SharePoint
• AWS S3
• SFTP
• SAP Build Work Zone
• SAP Document Management service
• ServiceNow
• Google Drive

Bruno collections are provided for AWS S3, Microsoft SharePoint, and SAP Build Work Zone (see the next subsection). For the other data repositories, refer to the SAP documentation:

• Orchestration (V2) with Grounding Capabilities in SAP AI Core | SAP Tutorials
• Generic Secrets for Grounding | SAP Help Portal
• Create a Document Grounding Pipeline Using the Pipelines API | SAP Help Portal

3.1 e) Credentials Required for the Next Steps

Every Bruno collection expects the AI Core service key values. In addition, each data repository needs its own credentials and ships with its own collection.

AWS S3

Bruno collection: aicore-dg-s3.yml

• S3_Region, S3_Host, S3_Bucket, S3_Username, S3_Access_Key_ID, S3_Secret_Access_Key

Note: This collection works with any AWS S3 source. When using SAP BTP Object Store (as in the Bruno and AI Launchpad subsections), the values are taken from the Object Store service key.

Microsoft SharePoint

Bruno collection: aicore-dg-sharepoint.yml

• Microsoft Entra application: MS_Entra_Tenant_ID, MS_Entra_Application_ID, MS_Entra_Application_Client_Secret
• SharePoint site: SharePoint_URL

SAP Build Work Zone

Bruno collection: aicore-dg-workzone.yml

• Advanced Work Zone: DWS_URL, OAuthClient_Key, OAuthClient_Secret

3.1 f) Test Documents

Sample documents are provided to verify the end-to-end setup. Each file contains fictional content written with unique terminology, and the matching test prompt is included on the first page of the document so you can copy it straight into SAP Joule for Consultants. Upload at least one to your data repository:

• Test_document–Aurora_Phase_Conductor.pdf
• Test_document–Heliotrope_Fabricator.pdf
• Test_document–Obsidian_Flux_Harmonizer.pdf
• Test_document–Verdant_Chronoweaver.pdf

3.2 AI Core instance creation

3.2 a) Verify the AI Core Entitlement

Before running the booster, confirm that the extended plan of SAP AI Core is assigned to your global account and added to the target subaccount. The booster will fail if the entitlement is missing at the global account level.

• At the global account level, open Entitlements → Service Assignments.
• Search for SAP AI Core and confirm the extended plan is assigned.

3.2 b) Launch the AI Core Booster

We shall now execute the AI Core Booster to provision SAP AI Core in your subaccount. The booster automates the entitlement assignment, instance creation, and service key generation for the extended plan required by AI Core Document Grounding.

• In your global account, click on Boosters, look for Set Up Account for SAP AI Core, and then click on Start to launch the booster.

• The booster checks that your user has the required authorizations and that the global account holds the necessary entitlements. Wait for both checks to show DONE, then click on Next Step.

• Select Select Subaccount to reuse an existing Cloud Foundry subaccount and click on Next Step.

Note: Choose Create Subaccount instead if you do not already have a target subaccount.

• Select your target Subaccount and Space. The Org is filled in automatically. Click on Next Step.

• Validate your selections and click on Finish.

• The booster assigns the extended plan, creates the default_aicore instance, and generates a default service key. Once completed, a Success message appears. Click on Navigate to Subaccount.

3.2 c) Download the AI Core Service Key

The booster creates the SAP AI Core instance and a service key automatically. Download the service key — you’ll need its credentials when configuring the Destination Service later.

• In the subaccount, open Instances and Subscriptions and click the default_aicore row to open its details panel on the right.
• Under Service Keys, click the … menu next to the default key and select Download.

3.3 AI Core Document Grounding with Bruno

The next two subsections — 3.3 AI Core Document Grounding with Bruno and 3.4 AI Core Document Grounding with AI Launchpad — describe two interchangeable paths to the same outcome. Pick one. Bruno is free and open-source; AI Launchpad is a UI-driven alternative that requires a paid subscription.

3.3 a) Install Bruno and Import the Collection

Bruno is an open-source API client for exploring and testing APIs. Use it to create the resource group, register the Object Store credentials as a generic secret, and provision the document grounding pipeline with the AI Core Document Grounding REST API.

• Download and install the latest version of Bruno.

Note: The templates collections are stored as .yml files. Older Bruno versions do not support the YAML import format, so make sure to install the latest release.

• Open Bruno and click the + icon next to Collections in the left navigation, then select Import collection.
• Pick the aicore-dg-s3.yml collection file when prompted.

3.3 b) Configure the Environment

The collection ships with an aicore-dg-s3 environment containing placeholders for the AI Core and Object Store credentials. Fill them in before running any request.

• With the aicore-dg-s3 collection open, click the environment selector in the top-right corner and select aicore-dg-s3, then click Configure.
• Replace each <placeholder> with the corresponding value:
  • AICore_AI_API_URL, AICore_clientid, AICore_clientsecret, AICore_url — from the AI Core service key.
  • S3_Region, S3_Host, S3_Bucket, S3_Username, S3_Access_Key_ID, S3_Secret_Access_Key — from the Object Store service key.
• Leave resource_group and generic_secret at their defaults (or adjust to your naming convention). For the Object Store setup, you can use object-store as the name for the generic secret. Leave pipeline_id empty for now.
• Click Save.

3.3 c) Fetch an Access Token

Every request to the AI Core API requires an OAuth2 bearer token. The collection retrieves it once and reuses it across the other requests via a collection-level script.

• Open the 01_get_token folder and select the get_token request.
• Click Send. A 200 OK response with an access_token confirms the credentials are correct.

Note: Occasionally, the token expires. Resend the request above to retrieve a new token.

3.3 d) Create the Resource Group

The resource group isolates document grounding artifacts inside the AI Core tenant. It must carry the ext.ai.sap.com/document-grounding: true label so that AI Core enables the grounding capability for it.

• Open the 02_resource_group folder and select create_resource_group.
• Click Send. A 202 Accepted response with the resourceGroupId confirms the request was queued.

• Select get_resource_group and click Send to poll the provisioning status. Initially the status field reads PROVISIONING.
• Wait a few minutes and resend the request until status reads PROVISIONED.

Note: Do not continue to the next step until the resource group is fully PROVISIONED — the generic secret cannot be created against a resource group that is still onboarding.

3.3 e) Create the Generic Secret

The generic secret stores the S3 credentials inside AI Core so that the document grounding pipeline can read from the Object Store bucket. The ext.ai.sap.com/documentRepositoryType: S3 label tells AI Core how to interpret the credentials.

• Open the 03_generic_secret folder and select create_generic_secret.
• Click Send. A 200 OK response with "message": "secret has been created" confirms the secret is registered.

3.3 f) Create the Pipeline

The pipeline binds the generic secret to a document grounding job. Once created, AI Core ingests, chunks, embeds, and indexes every document in the bucket.

Note: The full pipeline API is documented in Create a Document Grounding Pipeline Using the Pipelines API | SAP Help Portal. There you can find information on how to restrict processing to specific folders, include metadata, and configure the update schedule.

A request body that limits indexing to a folder inside the bucket looks like this (S3 only — for the other data repositories, see the SAP Help link above):

{
“type”: “S3”,
“configuration”: {
“destination”: “{{generic_secret}}”,
“s3”: {
“includePaths”: [
“/j4c-custom-knowledge”
]
}
}
}

To run the example above, replace the body in the create_pipeline request with the snippet.

• Open the 04_pipeline folder and select create_pipeline.
• Click Send. A 201 Created response returns the pipelineId.

• Note down the returned pipelineId for later use.
• Select get_pipeline_status and click Send to monitor progress. The pipeline takes a few minutes to finish — longer if more documents are in the bucket. Resend until status reads FINISHED.

• Select get_documents and click Send to list the documents picked up by the pipeline. Each document should report "status": "INDEXED".

Note: If you upload new documents to the Object Store bucket after the pipeline has finished, run the restart_pipeline request to trigger a fresh sync. The pipeline picks up additions and changes, then transitions back to FINISHED once the new documents are indexed.

3.4 AI Core Document Grounding with AI Launchpad

This subsection is an alternative to 3.3 AI Core Document Grounding with Bruno — pick one. The two paths produce the same resource group, generic secret, and pipeline. AI Launchpad is the UI-driven option.

3.4 a) Provision SAP AI Launchpad

SAP AI Launchpad is a web-based UI that lets you set up AI Core Document Grounding through a graphical interface — without writing API requests. Using it is optional: the same setup can be completed with Bruno (see 3.3 AI Core Document Grounding with Bruno). Choose AI Launchpad if you prefer a UI-driven workflow.

SAP AI Launchpad is a paid subscription service. The standard plan is required. Review the pricing on SAP Discovery Center Service – SAP AI Launchpad before subscribing.

SAP AI Launchpad can be set up using the Set Up Account for SAP AI Launchpad booster. Follow steps similar to the AI Core booster execution described in 3.2 AI Core instance creation.

Access to SAP AI Launchpad and its administrative features requires specific role collections to be assigned to your user. For more information, see SAP AI Launchpad > Roles and Authorizations | SAP Help Portal.

3.4 b) Create the AI API Connection

Before SAP AI Launchpad can manage the AI Core tenant, register the AI Core service key as an AI API connection.

• In SAP AI Launchpad, open Workspaces in the left navigation and click Add in the top-right corner.
• Enter a Connection Name (e.g., j4c-ai-core).
• Next to Service Key, click the upload icon and select the AI Core service key file you downloaded earlier. The AI API URL, XSUAA URL, Client ID, and Client Secret fields are populated automatically.
• Click Create.

3.4 c) Create the Resource Group

The resource group isolates document grounding artifacts inside the AI Core tenant.

• Go to SAP AI Core Administration > Resource Groups and click Create in the top-right corner.
• Enter a Resource Group ID (e.g., j4c-custom-knowledge).
• Ensure that the ext.ai.sap.com/document-grounding label is enabled.
• Click Create.

Note: The resource group takes a few minutes to transition from PROVISIONING to PROVISIONED. Do not continue to the next step until the resource group is fully provisioned — the generic secret cannot be created against a resource group that is still onboarding.

• In the left navigation, click Workspaces.
• Under Resource Groups, click the card for the resource group you created (e.g., j4c-custom-knowledge). The card should show Document Grounding: Enabled.
• The header at the top of the launchpad now displays the active workspace as <connection> (<resource-group>) — for example, j4c-ai-core (j4c-custom-knowledge).

3.4 d) Create the Generic Secret

The generic secret stores the S3 credentials inside AI Core so that the document grounding pipeline can read from the Object Store bucket.

• Go to SAP AI Core Administration > Generic Secrets and click Add.
• Select Amazon S3 as the Document Repository Type and ensure that Document Grounding is toggled on.
• Fill in the Secret key/value pairs using the Object Store service key.
• Click Add.

3.4 e) Create the Document Grounding Pipeline

Create a pipeline so AI Core ingests, chunks, embeds, and indexes every document in the data repository.

• Go to Generative AI Hub > Grounding Management and click Create in the top-right corner.
• Choose S3 as the Document Store Type.
• Choose the generic secret you created in the previous step (e.g., object-store)
• Click Create.

The pipeline will now start ingesting the documents from the data repository. This process will take a few minutes. Open the pipeline to confirm that it finished successfully and that the test document was indexed.

• On the Grounding Management page, click the new pipeline (named pipeline-<id>-collection) to open its details.
• Wait until the Status in the header switches to Completed.
• Under Data Repository Content, the Documents list shows every indexed file.

Note: If you upload new documents to the Object Store bucket after the pipeline has finished, you can click the Synchronize icon in the top-right corner of the pipeline page to trigger a resync.

3.5 Support

3.5 a) Where to Open Support Tickets

Custom Knowledge Grounding spans multiple SAP services, and each layer is owned by a different team. When something goes wrong, route the ticket to the component that owns the layer where the issue occurs.

3.5 b) AI Core Document Grounding

The resource group, generic secret, and pipeline are part of AI Core Document Grounding. Open a ticket under component CA-ML-RAGE for issues such as:

• Resource group stuck in PROVISIONING or failing to provision.
• Generic secret creation rejected.
• Pipeline not starting, hanging, or finishing with errors.
• Documents in the data repository not indexed properly.

Note: When the SAP Joule for Consultants Console shows a pipeline as Completed With Errors, only the successfully indexed documents are used in answers. Inspect the per-document status in Bruno (via get_documents) or in SAP AI Launchpad before opening the ticket.

3.5 c) Work Zone as a Data Repository

If you use SAP Build Work Zone instead of Object Store as the data repository for document grounding, issues that are specific to the Work Zone source belong to component EP-WZ-STA.

Note: Issues with the AI Core pipeline that reads from Work Zone (indexing, embeddings, pipeline status) still belong under CA-ML-RAGE. Use EP-WZ-STA only when the problem is on the Work Zone side of the integration.

3.5 d) Destination Service and SAP Joule for Consultants Console

Once the pipeline reports FINISHED (Bruno) or Completed (AI Launchpad), the remaining steps described in the following sections — the setup of the Destination Service instance and the configuration in the SAP Joule for Consultants Console — is owned by the SAP Joule for Consultants team. Open a ticket under component CA-GAIF-SCC in this case.

Note: Before opening a CA-GAIF-SCC ticket, confirm that the pipeline itself is healthy. A failing pipeline surfaces in the console as a downstream symptom, but the root cause belongs to AI Core Document Grounding.

4. Connection to SAP Joule for Consultants

4.1 Destination Service setup

4.1 a) Configure the Destination Service

A Destination Service connects AI Core Document Grounding to SAP Joule for Consultants. Configure it as follows.

• In your subaccount, open Instances and Subscriptions and click Create.
• Select Destination Service with the lite plan, fill in the required fields, and click Create to provision the instance.

• Click the instance row (not the blue instance name) to open its details panel on the right.
• Under Service Keys, click Create.
• Enter a name and click Create.

• Download the service key — you’ll need it later to connect SAP Joule for Consultants.
• Then open the instance to configure its destinations — either click the blue instance name in the Instances list, or click Manage Instance at the top of the details panel on the right. Both navigate to the same instance page.

Note: Destinations can be configured either at the instance level or at the subaccount level. This setup uses the instance level. Do not configure the destination at the subaccount level.

Note: Editing or deleting destinations requires the BTP role collection Destination Administrator.

4.1 b) Create the Destination

• In the Destination Service instance, open the Destinations section and click Create.
• Select From Scratch and click Create.

• Open the service key of the AI Core instance — you’ll copy several values from it.
• Set Authentication to OAuth2ClientCredentials, then paste in the Client ID and Client Secret from the AI Core service key.
• Fill in the following fields, replacing the placeholders with the values from the service key:
  • Token Service URL: <url>/oauth/token
  • URL: <AI_API_URL>/v2/lm/document-grounding
• Add the following Additional Properties:
  • HTML5.ForwardAuthToken: true
  • HTML5.DynamicDestination: true
  • AI-Resource-Group: name of your resource group

Note: AI-Resource-Group is not available as a predefined value in the value help — enter the key name manually.

• Check the connection to verify the destination is configured correctly.

4.2 Admin Console configuration

4.2 a) Enable Custom Knowledge Grounding

With the Destination Service configured, enable Custom Knowledge Grounding in the SAP Joule for Consultants Console and connect to AI Core Document Grounding through the destination.

Note: You need to have admin access to the SAP Joule for Consultants Console, see SAP Joule for Consultants Console.

• In the SAP Joule for Consultants Console, open the Settings menu from the left navigation and select Custom Knowledge Grounding.
• Toggle the switch to enable the feature.

4.2 b) Connect to the SAP Destination Service

• In the left navigation, open the Custom Knowledge Grounding page.
• Click Connect to open the connection dialog.
• Select JSON as the credentials type. Alternatively, choose Credentials to enter each field individually.
• Paste the contents of the Destination Service service key downloaded earlier into the JSON Credentials field.
• Click Next.

Note: Use the service key from the Destination Service instance — not the AI Core service key.

Once connected, the console retrieves the document grounding instances available through the destination.

• In the Select Sources dialog, check the instance(s) you want to use. Only sources with a green checkmark can be selected.
• Use Recheck Health if you recently updated a source and want to refresh its validation status.
• Click Save.

4.2 c) Manage Sources & Pipelines

The selected sources appear under Sources & Pipelines. Expand a source to inspect its pipelines.

The pipeline table shows the following columns:

The three action icons in the top-right corner of the page are:

• Pencil — configure document grounding instances (connect or disconnect sources).
• Bin — delete the selected source.
• Refresh — re-read the pipeline metadata from AI Core Document Grounding and update what the console shows (status, last sync timestamp, document count).

Note: Once the pipeline Status shows Completed, SAP Joule for Consultants references the grounded knowledge starting with the next new conversation.

Note: If the status shows Completed With Errors, some documents were not ingested successfully by the document grounding pipeline. Only successfully processed documents are used in answers. For details, check the documents in Bruno or in the SAP AI Launchpad. For support, please open a ticket under component CA-ML-RAGE — troubleshooting the AI Core Document Grounding pipeline is out of scope for SAP Joule for Consultants.

Note: The Refresh icon in the console only refreshes the view — it does not re-index your data repository. Whether newly uploaded documents become available to SAP Joule for Consultants is decided by re-running the AI Core Document Grounding pipeline itself: the restart_pipeline request in 3.3 AI Core Document Grounding with Bruno, or the Synchronize icon on the pipeline page in 3.4 AI Core Document Grounding with AI Launchpad. Until that pipeline run reaches FINISHED / Completed, the new documents are not indexed.

4.2 d) Troubleshooting

Destination shows “not a valid DGS destination”: The destination is missing one or more of the required additional properties (HTML5.DynamicDestination, HTML5.ForwardAuthToken, AI-Resource-Group), or its URL does not point to a valid document grounding endpoint. Edit the destination in the BTP Cockpit, add the missing properties, save it, then click Recheck Health in the console.

Source not appearing in the Select Sources dialog: Verify that the Destination Service instance whose service key you connected is the same instance where you created the destination. The console only discovers destinations from its own bound Destination Service instance.

No pipelines appear after saving: Allow a few minutes for the pipeline to initialize and the first sync to complete. Use the Refresh icon on the Sources & Pipelines page to update the view.

5. Testing

5.1 Testing Custom Knowledge Grounding

5.1 a) Start a New Conversation

SAP Joule for Consultants references the grounded documents in any new conversation. Start a fresh conversation and ask a question whose answer is only available in the uploaded test document.

• In SAP Joule for Consultants, click New Conversation in the left navigation.
• Enter a prompt that targets content from the test document, here: Explain the Heliotrope Fabricator.

Note: The Heliotrope Fabricator is fictional content contained only in Test_document--Heliotrope_Fabricator.pdf. If SAP Joule for Consultants returns a meaningful explanation, the document grounding is working as expected.

5.1 b) Verify the Source Citation

SAP Joule for Consultants cites the documents it used to ground its answer. Open the sources panel to confirm that the test document was retrieved.

• Once SAP Joule for Consultants has answered, click Sources below the response.
• A panel opens on the right listing the cited documents.

The panel header reads Sources and lists every document the answer was grounded on. Entries provided through Custom Knowledge Grounding are grouped under the Customer category, with each entry labeled Customer Document to distinguish it from SAP-managed content. The cited file is shown with the citation index from the answer (e.g., [1]) and a page reference (e.g., (pg 1)).

Note: Source entries are clickable, but for S3-backed sources the link does not resolve to the original file. Because the Object Store uses S3, this applies to all documents uploaded through it.

“}]] 

 [[{“value”:”Enhance SAP Joule for Consultants with Custom Knowledge GroundingTable of Contents1. Introduction1.1 Overview1.2 Recommendations2. Data Repositories2.1 Data Repositories – Overview2.2 Object Store2.3 Microsoft SharePoint2.4 SAP Build Work Zone3. AI Core Document Grounding3.1 AI Core Document Grounding – Overview3.2 AI Core instance creation3.3 AI Core Document Grounding with Bruno3.4 AI Core Document Grounding with AI Launchpad3.5 Support4. Connection to SAP Joule for Consultants4.1 Destination Service setup4.2 Admin Console configuration5. Testing5.1 Testing Custom Knowledge Grounding1. Introduction1.1 Overview1.1 a) What Custom Knowledge Grounding DoesCustom Knowledge Grounding extends SAP Joule for Consultants with your organization’s own documents. The documents are uploaded to a data repository, indexed by SAP AI Core Document Grounding, and retrieved at runtime to ground SAP Joule for Consultants’ answers. Users get responses that combine the SAP-curated content SAP Joule for Consultants ships with and their firm’s own methodologies, policies, and delivery standards.1.1 b) PrerequisitesSAP Joule for Consultants — an existing instance. See Onboarding Guide for SAP Joule for Consultants | SAP Help Portal.SAP AI Core with the extended plan — required by Document Grounding. See SAP AI Core — SAP Discovery Center.Object Store (recommended for the initial setup) — the simplest data repository to start with, and the example used throughout this guide. See Object Store — SAP Discovery Center.1.1 c) PricingCustom Knowledge Grounding relies on paid SAP services that are billed separately from SAP Joule for Consultants:SAP AI Core (extended plan) is mandatory. Charges accrue per document indexed and per token retrieved — see the Generative AI Hub in SAP AI Core Calculator and Pricing.Object Store (standard plan) if you take the recommended setup path. See Pricing.SAP AI Launchpad (standard plan) is optional — needed only if you prefer the UI-driven setup over the REST API path. See Pricing.Review the pricing of each service before subscribing.1.1 d) What This Guide CoversThe remaining sections walk through the end-to-end setup:2. Data Repositories — pick a supported data repository (e.g., Object Store, Microsoft SharePoint, or SAP Build Work Zone) and upload your documents to it.3. AI Core Document Grounding — provision SAP AI Core with the extended plan, then create a pipeline that ingests, chunks, embeds, and indexes the documents.4. Connection to SAP Joule for Consultants — configure a Destination Service instance, then enable Custom Knowledge Grounding in the SAP Joule for Consultants Console and connect it to the pipeline.5. Testing — verify the setup by asking SAP Joule for Consultants a question whose answer is only available in your uploaded documents.Note: Depending on what is already in place — an existing AI Core Document Grounding pipeline, ready-to-use data repository credentials, or neither — you may be able to skip directly to a later section. The next section, 1.2 Recommendations, opens with a decision tree for choosing the right entry point and then covers data repository choice and subaccount layout.1.2 Recommendations1.2 a) Where to StartNot every reader has to work through every section of this guide. Use the decision tree below to pick the entry point that matches what you already have in place:An AI Core Document Grounding pipeline is already provisioned (resource group, generic secret, and a pipeline reporting FINISHED against the data repository you intend to use) → skip ahead to the 4. Connection to SAP Joule for Consultants section, starting with 4.1 Destination Service setup. Configure the Destination Service against the existing pipeline and enable Custom Knowledge Grounding in the SAP Joule for Consultants Console.The data repository credentials are already available (e.g., a SharePoint site with the Entra app registered, an S3 bucket, or a Work Zone tenant with Document Grounding enabled) and you have run an AI Core Document Grounding setup before → skip the 2. Data Repositories section and start with the 3. AI Core Document Grounding section.Neither of the above applies → start at the 2. Data Repositories section with 2.2 Object Store, as described below. It is the fastest path to a working setup and the one the rest of this guide uses as the running example.1.2 b) Start with Object StoreCustom Knowledge Grounding can read from any of the supported data repositories. For the initial setup, Object Store is the recommended choice:Fast activation — the Object Store instance is provisioned by a booster and ready in minutes. SharePoint, customer-managed S3, and Work Zone setups often take longer than anticipated because of onboarding complications, approval, and security processes.Very low cost — Object Store charges are small. Even when run for a few months until the target data repository is ready, the extra cost is negligible — see Pricing.Reversible — once the target data repository is online and indexed, point Custom Knowledge Grounding at the new pipeline and decommission the Object Store instance.The key advantage is that Custom Knowledge Grounding can be live in SAP Joule for Consultants while the longer-running setup of your target data repository runs in parallel — instead of blocking the entire rollout on it.Note: The Object Store instance must be provisioned in an AWS-based subaccount, since AI Core Document Grounding consumes it through the AWS S3 API.Note: Answers themselves are generated the same way regardless of the data repository. Source citations are clickable for SharePoint and Work Zone — they link back to the original file. For Object Store and the other repositories, citations show the document name and page reference but do not resolve to the file.1.2 c) Subaccount Layout and Existing InstancesThe four moving parts — Object Store (or any other data repository), SAP AI Core, Destination Service, and SAP Joule for Consultants — communicate through service-key credentials, which allows for very flexible placement: they can all sit in the same subaccount, in different subaccounts, or even in different global accounts.Reuse existing instances where available. If your organization already runs SAP AI Core (with the extended plan) or SAP AI Launchpad in another subaccount, you can reuse them instead of provisioning new instances.The Destination Service can be deployed in the SAP Joule for Consultants subaccount or in the AI Core subaccount. In general, it is recommended to place it in the same subaccount as the consuming application—in this case, the SAP Joule for Consultants subaccount.The Object Store must live in an AWS-based subaccount (see above). The SAP Joule for Consultants subaccount is a convenient option, since it runs on AWS data centers.2. Data Repositories2.1 Data Repositories – Overview2.1 a) What a Data Repository ProvidesAI Core Document Grounding does not host the source documents itself — it reads them from an external data repository and serves the resulting index to SAP Joule for Consultants. The data repository is the system of record where you upload and maintain the documents you want to ground SAP Joule for Consultants’ answers on.The starting point for the AI Core Document Grounding documentation is Grounding | SAP Help Portal.2.1 b) Supported Data RepositoriesAI Core Document Grounding can read content from the following data repositories:Microsoft SharePointAWS S3SFTPSAP Build Work ZoneSAP Document Management serviceServiceNowGoogle DriveSetup steps for Google Drive, Microsoft SharePoint, SAP Build Work Zone, and ServiceNow (SNOW) are described on Set Up Data Repository Integration | SAP Help Portal.Note: That help page is written for Document Grounding with SAP Joule for Business. Only the data repository setup instructions apply here — the remaining content on the page is specific to SAP Joule for Business and is not relevant for SAP Joule for Consultants.The next subsections walk through the setup of three data repositories in more detail:2.2 Object Store2.3 Microsoft SharePoint2.4 SAP Build Work ZoneThe three subsections are independent — pick the repository (or repositories) you want to use and skip the others. For any data repository not covered here, refer to the SAP Help pages linked above.2.2 Object Store2.2 a) Provision the Object Store ServiceThe Object Store provides S3-compatible storage where documents are uploaded for AI Core Document Grounding to index. Begin by verifying the entitlement at the global account and adding the service plan to your subaccount.Note: Create the Object Store in an AWS-based subaccount; the steps below assume an AWS backend.At the global account level, open Entitlements → Service Assignments.Search for Object Store and confirm the standard plan is assigned.In your subaccount, open Entitlements and click Edit.Click Add Service Plans, select Object Store, check the standard plan, and click Add 1 Service Plan.Click Save to apply the changes.2.2 b) Set Up the Cloud Foundry EnvironmentThe Object Store instance runs in the Cloud Foundry runtime, so the subaccount must have Cloud Foundry enabled and a space available to host the instance.In your subaccount, open Overview and click Enable Cloud Foundry.Keep the Cloud Foundry Runtime environment and standard plan, adjust the Instance Name and Org Name if needed, then click Create.Once Cloud Foundry is enabled, click Create Space on the Cloud Foundry Environment tab.Enter a space name (for example, j4c), assign yourself the Space Developer and Space Manager roles, and click Create.2.2 c) Create the Object Store InstanceIn your subaccount, open Instances and Subscriptions and click Create.Fill in the required fields:Service: Object StorePlan: standardRuntime Environment: Cloud FoundrySpace: the space you created (e.g., j4c)Instance Name: e.g., j4c-object-storeClick Create to provision the instance.Click on the instance row to open its details panel on the right.Under Service Keys, click Create.Enter a name (e.g., j4c-object-store-sk) and click Create.Once the key’s Status shows Created, click the … menu on the service key row and choose Download to save the credentials as a JSON file. You can also choose View to inspect them in the browser and copy individual values.Note: The service key contains the access_key_id, secret_access_key, region, and bucket values needed to upload documents — you’ll use them in the next step.2.2 d) Upload Documents with the AWS CLIWith the service key credentials, you can use the AWS CLI to upload documents to the Object Store bucket. AI Core Document Grounding will then index these documents for retrieval.Install the AWS CLI if it isn’t already on your machine.Note: If you prefer a graphical client, Cyberduck supports S3 with the same service key credentials.Download the sample file Test_document–Heliotrope_Fabricator.pdf. You can append –Object-Store to the filename (e.g., Test_document–Heliotrope_Fabricator–Object-Store.pdf) to mark which storage backend the copy belongs to.Run aws configure and paste in the credentials when prompted — the default output format can be left blank.Upload the document to the bucket and list its contents to verify:aws s3 cp <file> s3://<bucket>
aws s3 ls s3://<bucket>Note: The bucket name in the service key is prefixed with hcp- followed by a unique identifier.2.3 Microsoft SharePointDisclaimer: The steps below are a reference. Coordinate with your SharePoint administrator and validate against the documentation below.Set Up Data Repository Integration | SAP Help Portal — the official SAP guidance for this setup. It is written for SAP Joule for Business, but the data repository setup steps apply here as well.Understanding Resource Specific Consent for Microsoft Graph and SharePoint Online | Microsoft Learn — Microsoft’s reference on the Resource Specific Consent permission model that the steps below rely on.2.3 a) Register the Microsoft Entra ApplicationAI Core Document Grounding connects to a SharePoint site through a Microsoft Entra application using the OAuth2ClientCredentials flow. Start by registering the application in Microsoft Entra.Sign in to entra.microsoft.com with an account that can register applications.Open App registrations and click New registration.Enter a name (e.g., j4c-sharepoint-grounding), keep the default settings, and click Register.From the Overview tab, copy the Application (client) ID and the Directory (tenant) ID — you’ll need both for the 3. AI Core Document Grounding section.2.3 b) Grant the Sites.Selected PermissionThe application needs the Sites.Selected Microsoft Graph permission. On its own, this permission grants access to zero SharePoint sites — it only makes the application eligible to be granted access to specific sites in a later step. This is intentional and enforces least-privilege.In the registered application, open API permissions and click Add a permission.Select Microsoft Graph → Application permissions.Search for Sites.Selected, check it, and click Add permissions.Note: Use the Sites.Selected permission listed under Microsoft Graph — not the one with the same name listed under SharePoint. They are different resources, and AI Core Document Grounding consumes Microsoft Graph.Click Grant admin consent for [Your Org]. A tenant Global Administrator or Privileged Role Administrator must perform this step.Verify that the Status column for Sites.Selected shows a green check mark.2.3 c) Create a Client SecretThe OAuth2ClientCredentials flow authenticates with a client secret instead of a user sign-in.In the registered application, open Certificates & secrets and click New client secret.Enter a description, pick an expiry, and click Add.Copy the secret Value immediately — it is only displayed once. Store it together with the tenant ID and application ID.2.3 d) Link the Application to the SharePoint SiteAfter admin consent, the application still cannot read any site. You now have to bind it explicitly to the SharePoint site that hosts your grounding documents. The easiest path is the PnP.PowerShell module.Note: The account running the grant command must have the Site Collection Administrator role on the target SharePoint site (a tenant Global Administrator works as well). Owners of the Microsoft 365 Group or Teams team that backs a SharePoint site automatically have this role. If you don’t have it, ask a SharePoint administrator to grant it temporarily. Sites.Selected on its own is not enough to perform the grant.Install PowerShell 7.2 or later — see Install PowerShell on Windows, Linux, and macOS | Microsoft Learn.Install the PnP.PowerShell module — see Installing PnP PowerShell | PnP PowerShell.On first use in the tenant, register PnP’s own Entra application by running Register-PnPEntraIDApp and following the prompts. The resulting Client ID is the helper-app ID used to sign in with Connect-PnPOnline — it is not the ID of the application you registered in the previous sections.Note: Register-PnPEntraIDApp is a one-time setup per tenant. Skip it if PnP.PowerShell has already been initialized.Sign in to the target SharePoint site and grant the Sites.Selected application read access:Connect-PnPOnline -Url “<site-url>” -Interactive -ClientId “<pnp-helper-app-client-id>”
Grant-PnPAzureADAppSitePermission `
-AppId “<sites-selected-app-client-id>” `
-DisplayName “<display-name>” `
-Site “<site-url>” `
-Permissions ReadReplace <site-url> with the full SharePoint site URL (e.g., https://contoso.sharepoint.com/sites/KnowledgeBase), <pnp-helper-app-client-id> with the helper-app client ID from Register-PnPEntraIDApp, and <sites-selected-app-client-id> with the Application (client) ID of the application you registered for AI Core Document Grounding.A successful response prints the permission ID, the granted role (read), and the application identity.Note: Read is sufficient for AI Core Document Grounding. The same grant can also be issued through a Microsoft Graph POST /sites/{site-id}/permissions call — see Understanding Resource Specific Consent for Microsoft Graph and SharePoint Online | Microsoft Learn for the REST equivalent.2.3 e) Upload Documents to the SharePoint SiteOpen the SharePoint site in a browser and upload your documents to a document library.Download the sample file Test_document–Heliotrope_Fabricator.pdf. You can append –SharePoint to the filename (e.g., Test_document–Heliotrope_Fabricator–SharePoint.pdf) to mark which storage backend the copy belongs to.The Bruno and AI Launchpad subsections in the 3. AI Core Document Grounding section consume the following SharePoint credentials, which you have now collected:MS_Entra_Tenant_ID — Directory (tenant) ID from the Overview tab.MS_Entra_Application_ID — Application (client) ID from the Overview tab.MS_Entra_Application_Client_Secret — the secret value copied from Certificates & secrets.SharePoint_URL — the full URL of the SharePoint site you granted permission on.2.3 f) TroubleshootingThe Troubleshooting Document Grounding Pipelines | SAP Community blog walks setup failures down in three steps:Step 1 — Validate access to the Document Grounding API.Step 2 — Validate SharePoint credentials.Step 3 — Validate pipeline configuration.Especially noteworthy are the checks for step 3:SharePoint site identifier. Use the identifier from the site URL (e.g., HR%20Team for https://contoso.sharepoint.com/sites/HR%20Team), not the human-readable site title. Mind the URL encoding.Include paths. Match the URL path of the folder you want to ground on, relative to the document library — /teched, not /Shared%20Documents/teched.Pipeline created but no documents. Almost always means the include path is too narrow. Test with no include path first; once replication works, narrow it down.2.4 SAP Build Work Zone2.4 a) Provision SAP Build Work Zone, Advanced EditionDocument Grounding is available in SAP Build Work Zone, advanced edition across all plans. Provision the tenant if you don’t already have one.Follow the Onboarding to SAP Build Work Zone, advanced edition | SAP Help Portal guide.The community post SAP Build Work Zone: advanced edition — step-by-step configuration | SAP Community covers the same ground with screenshots and may be easier to follow on a first pass.2.4 b) Enable Document Grounding on the Work Zone TenantOnce the Work Zone tenant is up, enable Document Grounding so AI Core can ingest its content, and collect the OAuth credentials AI Core will use to read it.Follow Integration With Document Grounding | SAP Help Portal.The community blog Joule — Document Grounding with SAP Build Work Zone, advanced edition | SAP Community walks through the same enablement end-to-end.Note: Stop after Step 3 of the community blog. Step 4: Use an Existing Joule Subaccount or Create a New One to Set Up Document Grounding is replaced by the AI Core Document Grounding setup in the 3. AI Core Document Grounding section of this guide — the Destination Service and SAP Joule for Consultants Console steps later on perform the equivalent integration for SAP Joule for Consultants.2.4 c) Upload Documents to Work ZoneIn the Work Zone workspace configured for Document Grounding, upload your documents to the workspace folder.Download the sample file Test_document–Heliotrope_Fabricator.pdf. You can append –Work-Zone to the filename (e.g., Test_document–Heliotrope_Fabricator–Work-Zone.pdf) to mark which storage backend the copy belongs to.The Bruno and AI Launchpad subsections in the 3. AI Core Document Grounding section consume the following Work Zone credentials, which you collected during the Document Grounding enablement above:DWS_URL — the URL of the Work Zone tenant.OAuthClient_Key — the OAuth client key.OAuthClient_Secret — the OAuth client secret.2.4 d) SupportIssues specific to the Work Zone side of this setup — Document Grounding capability, OAuth configuration, content not appearing in the workspace folder — belong to support component EP-WZ-STA. Issues with the AI Core pipeline that reads from Work Zone (indexing, embeddings, pipeline status) belong under CA-ML-RAGE. The 3.5 Support subsection later in the next section has the full breakdown.3. AI Core Document Grounding3.1 AI Core Document Grounding – Overview3.1 a) What AI Core Document Grounding DoesSAP AI Core Document Grounding reads documents from a data repository, chunks them, generates embeddings, and stores them in a vector index. SAP Joule for Consultants queries that index at runtime to ground its answers in your organization’s own content.Note: For a conceptual introduction to what happens inside that pipeline — vector embeddings, the SAP HANA Cloud vector engine, and how data ingestion, orchestration, and retrieval fit together in the Generative AI Hub — see the Analyzing Document Grounding in Generative AI Hub | SAP Learning lesson. It is theoretical background, not setup instructions, and complements the hands-on subsections that follow.3.1 b) What the Rest of This Section CoversThe next subsections walk through the setup of SAP AI Core Document Grounding. The instructions in both the Bruno and AI Launchpad subsections use the Object Store as the example data repository:3.2 AI Core instance creation — provision SAP AI Core with the extended plan via the AI Core booster, and download the service key.3.3 AI Core Document Grounding with Bruno — create the resource group, generic secret, and pipeline through the REST API using an importable Bruno collection.3.4 AI Core Document Grounding with AI Launchpad — an alternative path that performs the same setup through a UI instead of the REST API.3.5 Support — where to open tickets when something goes wrong.Note: The Bruno and AI Launchpad subsections are two interchangeable paths to the same outcome. Bruno is a free, open-source API client; SAP AI Launchpad is an alternative UI-driven route that requires a paid subscription.3.1 c) PricingSAP AI Core offers a free plan, but Document Grounding requires the extended plan, which is a paid subscription. Charges accrue per document indexed and per token retrieved, and are billed separately from SAP Joule for Consultants. Review the pricing before subscribing:SAP AI Core — SAP Discovery CenterGenerative AI Hub in SAP AI Core CalculatorIf you take the AI Launchpad path, note that SAP AI Launchpad with the standard plan is also a paid subscription: SAP AI Launchpad — SAP Discovery Center.3.1 d) Supported Data RepositoriesAI Core Document Grounding can index content from the following data repositories:Microsoft SharePointAWS S3SFTPSAP Build Work ZoneSAP Document Management serviceServiceNowGoogle DriveBruno collections are provided for AWS S3, Microsoft SharePoint, and SAP Build Work Zone (see the next subsection). For the other data repositories, refer to the SAP documentation:Orchestration (V2) with Grounding Capabilities in SAP AI Core | SAP TutorialsGeneric Secrets for Grounding | SAP Help PortalCreate a Document Grounding Pipeline Using the Pipelines API | SAP Help Portal3.1 e) Credentials Required for the Next StepsEvery Bruno collection expects the AI Core service key values. In addition, each data repository needs its own credentials and ships with its own collection.AWS S3Bruno collection: aicore-dg-s3.ymlS3_Region, S3_Host, S3_Bucket, S3_Username, S3_Access_Key_ID, S3_Secret_Access_KeyNote: This collection works with any AWS S3 source. When using SAP BTP Object Store (as in the Bruno and AI Launchpad subsections), the values are taken from the Object Store service key.Microsoft SharePointBruno collection: aicore-dg-sharepoint.ymlMicrosoft Entra application: MS_Entra_Tenant_ID, MS_Entra_Application_ID, MS_Entra_Application_Client_SecretSharePoint site: SharePoint_URLSAP Build Work ZoneBruno collection: aicore-dg-workzone.ymlAdvanced Work Zone: DWS_URL, OAuthClient_Key, OAuthClient_Secret3.1 f) Test DocumentsSample documents are provided to verify the end-to-end setup. Each file contains fictional content written with unique terminology, and the matching test prompt is included on the first page of the document so you can copy it straight into SAP Joule for Consultants. Upload at least one to your data repository:Test_document–Aurora_Phase_Conductor.pdfTest_document–Heliotrope_Fabricator.pdfTest_document–Obsidian_Flux_Harmonizer.pdfTest_document–Verdant_Chronoweaver.pdf3.2 AI Core instance creation3.2 a) Verify the AI Core EntitlementBefore running the booster, confirm that the extended plan of SAP AI Core is assigned to your global account and added to the target subaccount. The booster will fail if the entitlement is missing at the global account level.At the global account level, open Entitlements → Service Assignments.Search for SAP AI Core and confirm the extended plan is assigned.3.2 b) Launch the AI Core BoosterWe shall now execute the AI Core Booster to provision SAP AI Core in your subaccount. The booster automates the entitlement assignment, instance creation, and service key generation for the extended plan required by AI Core Document Grounding.In your global account, click on Boosters, look for Set Up Account for SAP AI Core, and then click on Start to launch the booster.The booster checks that your user has the required authorizations and that the global account holds the necessary entitlements. Wait for both checks to show DONE, then click on Next Step.Select Select Subaccount to reuse an existing Cloud Foundry subaccount and click on Next Step.Note: Choose Create Subaccount instead if you do not already have a target subaccount.Select your target Subaccount and Space. The Org is filled in automatically. Click on Next Step.Validate your selections and click on Finish.The booster assigns the extended plan, creates the default_aicore instance, and generates a default service key. Once completed, a Success message appears. Click on Navigate to Subaccount.3.2 c) Download the AI Core Service KeyThe booster creates the SAP AI Core instance and a service key automatically. Download the service key — you’ll need its credentials when configuring the Destination Service later.In the subaccount, open Instances and Subscriptions and click the default_aicore row to open its details panel on the right.Under Service Keys, click the … menu next to the default key and select Download.3.3 AI Core Document Grounding with BrunoThe next two subsections — 3.3 AI Core Document Grounding with Bruno and 3.4 AI Core Document Grounding with AI Launchpad — describe two interchangeable paths to the same outcome. Pick one. Bruno is free and open-source; AI Launchpad is a UI-driven alternative that requires a paid subscription.3.3 a) Install Bruno and Import the CollectionBruno is an open-source API client for exploring and testing APIs. Use it to create the resource group, register the Object Store credentials as a generic secret, and provision the document grounding pipeline with the AI Core Document Grounding REST API.Download and install the latest version of Bruno.Note: The templates collections are stored as .yml files. Older Bruno versions do not support the YAML import format, so make sure to install the latest release.Open Bruno and click the + icon next to Collections in the left navigation, then select Import collection.Pick the aicore-dg-s3.yml collection file when prompted.3.3 b) Configure the EnvironmentThe collection ships with an aicore-dg-s3 environment containing placeholders for the AI Core and Object Store credentials. Fill them in before running any request.With the aicore-dg-s3 collection open, click the environment selector in the top-right corner and select aicore-dg-s3, then click Configure.Replace each <placeholder> with the corresponding value:AICore_AI_API_URL, AICore_clientid, AICore_clientsecret, AICore_url — from the AI Core service key.S3_Region, S3_Host, S3_Bucket, S3_Username, S3_Access_Key_ID, S3_Secret_Access_Key — from the Object Store service key.Leave resource_group and generic_secret at their defaults (or adjust to your naming convention). For the Object Store setup, you can use object-store as the name for the generic secret. Leave pipeline_id empty for now.Click Save.3.3 c) Fetch an Access TokenEvery request to the AI Core API requires an OAuth2 bearer token. The collection retrieves it once and reuses it across the other requests via a collection-level script.Open the 01_get_token folder and select the get_token request.Click Send. A 200 OK response with an access_token confirms the credentials are correct.Note: Occasionally, the token expires. Resend the request above to retrieve a new token.3.3 d) Create the Resource GroupThe resource group isolates document grounding artifacts inside the AI Core tenant. It must carry the ext.ai.sap.com/document-grounding: true label so that AI Core enables the grounding capability for it.Open the 02_resource_group folder and select create_resource_group.Click Send. A 202 Accepted response with the resourceGroupId confirms the request was queued.Select get_resource_group and click Send to poll the provisioning status. Initially the status field reads PROVISIONING.Wait a few minutes and resend the request until status reads PROVISIONED.Note: Do not continue to the next step until the resource group is fully PROVISIONED — the generic secret cannot be created against a resource group that is still onboarding.3.3 e) Create the Generic SecretThe generic secret stores the S3 credentials inside AI Core so that the document grounding pipeline can read from the Object Store bucket. The ext.ai.sap.com/documentRepositoryType: S3 label tells AI Core how to interpret the credentials.Open the 03_generic_secret folder and select create_generic_secret.Click Send. A 200 OK response with “message”: “secret has been created” confirms the secret is registered.3.3 f) Create the PipelineThe pipeline binds the generic secret to a document grounding job. Once created, AI Core ingests, chunks, embeds, and indexes every document in the bucket.Note: The full pipeline API is documented in Create a Document Grounding Pipeline Using the Pipelines API | SAP Help Portal. There you can find information on how to restrict processing to specific folders, include metadata, and configure the update schedule.A request body that limits indexing to a folder inside the bucket looks like this (S3 only — for the other data repositories, see the SAP Help link above):{
“type”: “S3”,
“configuration”: {
“destination”: “{{generic_secret}}”,
“s3”: {
“includePaths”: [
“/j4c-custom-knowledge”
]
}
}
}To run the example above, replace the body in the create_pipeline request with the snippet.Open the 04_pipeline folder and select create_pipeline.Click Send. A 201 Created response returns the pipelineId.Note down the returned pipelineId for later use.Select get_pipeline_status and click Send to monitor progress. The pipeline takes a few minutes to finish — longer if more documents are in the bucket. Resend until status reads FINISHED.Select get_documents and click Send to list the documents picked up by the pipeline. Each document should report “status”: “INDEXED”.Note: If you upload new documents to the Object Store bucket after the pipeline has finished, run the restart_pipeline request to trigger a fresh sync. The pipeline picks up additions and changes, then transitions back to FINISHED once the new documents are indexed.3.4 AI Core Document Grounding with AI LaunchpadThis subsection is an alternative to 3.3 AI Core Document Grounding with Bruno — pick one. The two paths produce the same resource group, generic secret, and pipeline. AI Launchpad is the UI-driven option.3.4 a) Provision SAP AI LaunchpadSAP AI Launchpad is a web-based UI that lets you set up AI Core Document Grounding through a graphical interface — without writing API requests. Using it is optional: the same setup can be completed with Bruno (see 3.3 AI Core Document Grounding with Bruno). Choose AI Launchpad if you prefer a UI-driven workflow.SAP AI Launchpad is a paid subscription service. The standard plan is required. Review the pricing on SAP Discovery Center Service – SAP AI Launchpad before subscribing.SAP AI Launchpad can be set up using the Set Up Account for SAP AI Launchpad booster. Follow steps similar to the AI Core booster execution described in 3.2 AI Core instance creation.Access to SAP AI Launchpad and its administrative features requires specific role collections to be assigned to your user. For more information, see SAP AI Launchpad > Roles and Authorizations | SAP Help Portal.3.4 b) Create the AI API ConnectionBefore SAP AI Launchpad can manage the AI Core tenant, register the AI Core service key as an AI API connection.In SAP AI Launchpad, open Workspaces in the left navigation and click Add in the top-right corner.Enter a Connection Name (e.g., j4c-ai-core).Next to Service Key, click the upload icon and select the AI Core service key file you downloaded earlier. The AI API URL, XSUAA URL, Client ID, and Client Secret fields are populated automatically.Click Create.3.4 c) Create the Resource GroupThe resource group isolates document grounding artifacts inside the AI Core tenant.Go to SAP AI Core Administration > Resource Groups and click Create in the top-right corner.Enter a Resource Group ID (e.g., j4c-custom-knowledge).Ensure that the ext.ai.sap.com/document-grounding label is enabled.Click Create.Note: The resource group takes a few minutes to transition from PROVISIONING to PROVISIONED. Do not continue to the next step until the resource group is fully provisioned — the generic secret cannot be created against a resource group that is still onboarding.In the left navigation, click Workspaces.Under Resource Groups, click the card for the resource group you created (e.g., j4c-custom-knowledge). The card should show Document Grounding: Enabled.The header at the top of the launchpad now displays the active workspace as <connection> (<resource-group>) — for example, j4c-ai-core (j4c-custom-knowledge).3.4 d) Create the Generic SecretThe generic secret stores the S3 credentials inside AI Core so that the document grounding pipeline can read from the Object Store bucket.Go to SAP AI Core Administration > Generic Secrets and click Add.Select Amazon S3 as the Document Repository Type and ensure that Document Grounding is toggled on.Fill in the Secret key/value pairs using the Object Store service key.Click Add.3.4 e) Create the Document Grounding PipelineCreate a pipeline so AI Core ingests, chunks, embeds, and indexes every document in the data repository.Go to Generative AI Hub > Grounding Management and click Create in the top-right corner.Choose S3 as the Document Store Type.Choose the generic secret you created in the previous step (e.g., object-store)Click Create.The pipeline will now start ingesting the documents from the data repository. This process will take a few minutes. Open the pipeline to confirm that it finished successfully and that the test document was indexed.On the Grounding Management page, click the new pipeline (named pipeline-<id>-collection) to open its details.Wait until the Status in the header switches to Completed.Under Data Repository Content, the Documents list shows every indexed file.Note: If you upload new documents to the Object Store bucket after the pipeline has finished, you can click the Synchronize icon in the top-right corner of the pipeline page to trigger a resync.3.5 Support3.5 a) Where to Open Support TicketsCustom Knowledge Grounding spans multiple SAP services, and each layer is owned by a different team. When something goes wrong, route the ticket to the component that owns the layer where the issue occurs.3.5 b) AI Core Document GroundingThe resource group, generic secret, and pipeline are part of AI Core Document Grounding. Open a ticket under component CA-ML-RAGE for issues such as:Resource group stuck in PROVISIONING or failing to provision.Generic secret creation rejected.Pipeline not starting, hanging, or finishing with errors.Documents in the data repository not indexed properly.Note: When the SAP Joule for Consultants Console shows a pipeline as Completed With Errors, only the successfully indexed documents are used in answers. Inspect the per-document status in Bruno (via get_documents) or in SAP AI Launchpad before opening the ticket.3.5 c) Work Zone as a Data RepositoryIf you use SAP Build Work Zone instead of Object Store as the data repository for document grounding, issues that are specific to the Work Zone source belong to component EP-WZ-STA.Note: Issues with the AI Core pipeline that reads from Work Zone (indexing, embeddings, pipeline status) still belong under CA-ML-RAGE. Use EP-WZ-STA only when the problem is on the Work Zone side of the integration.3.5 d) Destination Service and SAP Joule for Consultants ConsoleOnce the pipeline reports FINISHED (Bruno) or Completed (AI Launchpad), the remaining steps described in the following sections — the setup of the Destination Service instance and the configuration in the SAP Joule for Consultants Console — is owned by the SAP Joule for Consultants team. Open a ticket under component CA-GAIF-SCC in this case.Note: Before opening a CA-GAIF-SCC ticket, confirm that the pipeline itself is healthy. A failing pipeline surfaces in the console as a downstream symptom, but the root cause belongs to AI Core Document Grounding.4. Connection to SAP Joule for Consultants4.1 Destination Service setup4.1 a) Configure the Destination ServiceA Destination Service connects AI Core Document Grounding to SAP Joule for Consultants. Configure it as follows.In your subaccount, open Instances and Subscriptions and click Create.Select Destination Service with the lite plan, fill in the required fields, and click Create to provision the instance.Click the instance row (not the blue instance name) to open its details panel on the right.Under Service Keys, click Create.Enter a name and click Create.Download the service key — you’ll need it later to connect SAP Joule for Consultants.Then open the instance to configure its destinations — either click the blue instance name in the Instances list, or click Manage Instance at the top of the details panel on the right. Both navigate to the same instance page.Note: Destinations can be configured either at the instance level or at the subaccount level. This setup uses the instance level. Do not configure the destination at the subaccount level.Note: Editing or deleting destinations requires the BTP role collection Destination Administrator.4.1 b) Create the DestinationIn the Destination Service instance, open the Destinations section and click Create.Select From Scratch and click Create.Open the service key of the AI Core instance — you’ll copy several values from it.Set Authentication to OAuth2ClientCredentials, then paste in the Client ID and Client Secret from the AI Core service key.Fill in the following fields, replacing the placeholders with the values from the service key:Token Service URL: <url>/oauth/tokenURL: <AI_API_URL>/v2/lm/document-groundingAdd the following Additional Properties:HTML5.ForwardAuthToken: trueHTML5.DynamicDestination: trueAI-Resource-Group: name of your resource groupNote: AI-Resource-Group is not available as a predefined value in the value help — enter the key name manually.Check the connection to verify the destination is configured correctly.4.2 Admin Console configuration4.2 a) Enable Custom Knowledge GroundingWith the Destination Service configured, enable Custom Knowledge Grounding in the SAP Joule for Consultants Console and connect to AI Core Document Grounding through the destination.Note: You need to have admin access to the SAP Joule for Consultants Console, see SAP Joule for Consultants Console.In the SAP Joule for Consultants Console, open the Settings menu from the left navigation and select Custom Knowledge Grounding.Toggle the switch to enable the feature.4.2 b) Connect to the SAP Destination ServiceIn the left navigation, open the Custom Knowledge Grounding page.Click Connect to open the connection dialog.Select JSON as the credentials type. Alternatively, choose Credentials to enter each field individually.Paste the contents of the Destination Service service key downloaded earlier into the JSON Credentials field.Click Next.Note: Use the service key from the Destination Service instance — not the AI Core service key.Once connected, the console retrieves the document grounding instances available through the destination.In the Select Sources dialog, check the instance(s) you want to use. Only sources with a green checkmark can be selected.Use Recheck Health if you recently updated a source and want to refresh its validation status.Click Save.4.2 c) Manage Sources & PipelinesThe selected sources appear under Sources & Pipelines. Expand a source to inspect its pipelines.The pipeline table shows the following columns:ColumnDescriptionPipeline IDUnique identifier — click the link to open detailed processing logsIndexed DocumentsTotal number of documents submitted for processingData SourceType of connected data source (e.g., S3)Last SyncTimestamp of the most recent sync runStatusCompleted, Completed With Errors, or ProcessingThe three action icons in the top-right corner of the page are:Pencil — configure document grounding instances (connect or disconnect sources).Bin — delete the selected source.Refresh — re-read the pipeline metadata from AI Core Document Grounding and update what the console shows (status, last sync timestamp, document count).Note: Once the pipeline Status shows Completed, SAP Joule for Consultants references the grounded knowledge starting with the next new conversation.Note: If the status shows Completed With Errors, some documents were not ingested successfully by the document grounding pipeline. Only successfully processed documents are used in answers. For details, check the documents in Bruno or in the SAP AI Launchpad. For support, please open a ticket under component CA-ML-RAGE — troubleshooting the AI Core Document Grounding pipeline is out of scope for SAP Joule for Consultants.Note: The Refresh icon in the console only refreshes the view — it does not re-index your data repository. Whether newly uploaded documents become available to SAP Joule for Consultants is decided by re-running the AI Core Document Grounding pipeline itself: the restart_pipeline request in 3.3 AI Core Document Grounding with Bruno, or the Synchronize icon on the pipeline page in 3.4 AI Core Document Grounding with AI Launchpad. Until that pipeline run reaches FINISHED / Completed, the new documents are not indexed.4.2 d) TroubleshootingDestination shows “not a valid DGS destination”: The destination is missing one or more of the required additional properties (HTML5.DynamicDestination, HTML5.ForwardAuthToken, AI-Resource-Group), or its URL does not point to a valid document grounding endpoint. Edit the destination in the BTP Cockpit, add the missing properties, save it, then click Recheck Health in the console.Source not appearing in the Select Sources dialog: Verify that the Destination Service instance whose service key you connected is the same instance where you created the destination. The console only discovers destinations from its own bound Destination Service instance.No pipelines appear after saving: Allow a few minutes for the pipeline to initialize and the first sync to complete. Use the Refresh icon on the Sources & Pipelines page to update the view.5. Testing5.1 Testing Custom Knowledge Grounding5.1 a) Start a New ConversationSAP Joule for Consultants references the grounded documents in any new conversation. Start a fresh conversation and ask a question whose answer is only available in the uploaded test document.In SAP Joule for Consultants, click New Conversation in the left navigation.Enter a prompt that targets content from the test document, here: Explain the Heliotrope Fabricator.Note: The Heliotrope Fabricator is fictional content contained only in Test_document–Heliotrope_Fabricator.pdf. If SAP Joule for Consultants returns a meaningful explanation, the document grounding is working as expected.5.1 b) Verify the Source CitationSAP Joule for Consultants cites the documents it used to ground its answer. Open the sources panel to confirm that the test document was retrieved.Once SAP Joule for Consultants has answered, click Sources below the response.A panel opens on the right listing the cited documents.The panel header reads Sources and lists every document the answer was grounded on. Entries provided through Custom Knowledge Grounding are grouped under the Customer category, with each entry labeled Customer Document to distinguish it from SAP-managed content. The cited file is shown with the citation index from the answer (e.g., [1]) and a page reference (e.g., (pg 1)).Note: Source entries are clickable, but for S3-backed sources the link does not resolve to the original file. Because the Object Store uses S3, this applies to all documents uploaded through it.”}]] Read More Technology Blog Posts by SAP articles 

#SAPCHANNEL