Document Grounding Vector Store (the ‘AI Foundation Vector Store’)

Name check first

The phrase ‘AI Foundation Vector Store’ used on earlier versions of this card is not an SAP product name. SAP AI Foundation is SAP's umbrella for its AI platform services (SAP AI Core, the generative AI hub, SAP AI Launchpad, AI services such as SAP Document AI), now part of the SAP Business AI Platform announced at Sapphire in May 2026. The managed vector store that SAP actually documents sits inside the document grounding capability of the generative AI hub: its Pipelines API "fetches documents from a supported data source, preprocesses and chunks the document content, and generates semantic embeddings" which it "stores into the HANA Vector Store". This card describes that component. Building your own vector tables directly on SAP HANA Cloud's Vector Engine is a different design, covered in M330.

How grounding works end to end

Grounding is retrieval-augmented generation (RAG, C027): documents are split into chunks, each chunk is turned into an embedding, embeddings are stored; at question time the query is embedded, similar chunks are retrieved and passed to the model as context. SAP exposes three APIs for this, all under $AI_API_URL/v2/lm/document-grounding and all scoped to a resource group that carries the label ext.ai.sap.com/document-grounding = true.

Why it matters

• Joule's document grounding makes every ingested document available to all Joule users of the tenant and does not filter answers by user permissions — ingesting confidential content is a data leak by design.
• Managed pipelines refresh daily and hold at most 8,000 documents each; sizing and freshness expectations must be set before the pilot, not after.
• Naming a non-existent ‘AI Foundation Vector Store’ in a proposal hides the real choice between managed document grounding, HANA Cloud Vector Engine and the Knowledge Graph.

Key points

• No SAP product is called ‘AI Foundation Vector Store’; the documented managed store is the SAP HANA vector store behind generative AI hub document grounding.
• Three APIs under /v2/lm/document-grounding: Pipelines (repository sync), Vector (collections/documents/chunks you upload), Retrieval (search, merge and rank).
• Resource group must carry the label ext.ai.sap.com/document-grounding = true; repositories are registered through generic secrets.
• Sources: SharePoint, AWS S3, SFTP, SAP Build Work Zone, SAP Document Management, ServiceNow, Google Drive; daily refresh; max 8,000 documents per pipeline.
• Quotas: 55,000 collections, 100,000 docs/collection, 200,000 chunks/doc; searches return max 100 chunks or 10 documents; 250 vector searches/min.
• Metering: grounding storage (GB-day) and retrieval (text blocks) converted into capacity units, separate from model tokens.
• Joule document grounding: ingested documents visible to all Joule users of the tenant; no user-permission filtering yet.
• Alternatives: HANA Cloud Vector Engine directly (own chunking, row-level security) or SAP Knowledge Graph for entity relationships.

Terms used on this page

Document grounding

Generative AI hub capability that ingests, embeds, stores and retrieves enterprise documents to ground model answers (RAG).

Pipelines API

Creates pipelines that fetch documents from a registered repository, chunk and embed them and store them in the HANA vector store.

Vector API

Lets you create collections and upload documents and chunks directly, with metadata at every level.

Retrieval API

Similarity search over repositories and collections, with filters, merge and ranking of results.

Collection

Logical container of embedded documents in the Vector API.

Chunk

Small text segment of a document; the unit that is embedded and retrieved.

Generic secret (grounding)

Stored credentials that register a repository such as SharePoint, S3 or SFTP for a grounding pipeline.

Grounding Management

SAP AI Launchpad app to create repositories, trigger syncs, preview chunks and run test searches.

Sources

1. SAP AI Core — Grounding (SAP-docs)
2. SAP AI Core — Pipelines API (SAP-docs)
3. SAP AI Core — Vector API, quotas and limits (SAP-docs)
4. SAP AI Core — Data Management APIs, rate limits (SAP-docs)
5. SAP AI Core — Create a Resource Group for Grounding (SAP-docs)
6. SAP AI Launchpad — Grounding Management (SAP-docs)
7. Integrating Joule with SAP Solutions — Set Up Document Grounding (PDF, 14 Sep 2026)
8. SAP Community — Grounding Joule Agent on Your Own Documents Without Object Store (11 Sep 2026)
9. SAP Learning — Describing SAP AI Foundation
10. SAP Developers — Tutorial: orchestration service document grounding (accessed 2026-09)
11. SAP Help — Orchestration service (generative AI hub) overview
12. SAP Help — HANA Cloud Vector Engine: REAL_VECTOR/HALF_VECTOR data types (build-your-own alternative, M330)
13. SAP Community — Knowledge graphs for LLM grounding and avoiding hallucination