Artificial intelligence has long since established itself as a driving force, the entire industries circulated. In many cases, however, in the IT landscapes of companies there is still a lack of uniform AI processes: some services use vector searches, some use relational databases and other work with knowledge graphs. With each additional component, complexity, latency and costs increase.

This is exactly why it is time to ask yourself what a modern AI-capable database should look like.

And SAP HANA Cloud offers an answer: a uniform multi-model platform, the vector, diagram, text, geo- and relational data brings together. This approach enables developers to create more intelligent AI solutions with even more contextual reference based on operational data.

Uniform database for all models enables complex AI workloads

SAP HANA Cloud offers the following:

• Vector data For semantic search and similarity searches
• Diagram data For the representation of clear connections and knowledge diagrams
• Text and geodata For practical context
• Relational data For structured processes and analyzes

Instead of transmitting data across various services, all data can be saved and processed at one point. This enables faster added value and at the same time reduces the risk of deviations.

In this way, the multi-model approach can be ideally implemented and the foundation for powerful, scalable AI workloads can be placed.

Semantics and similarity: combine vector searches with knowledge graphs

Traditional semantic search engines can indicatewhich Documents are similar, but you cannot justify Why This is so. Knowledge graphs, in turn, can represent clear connections in detail, but the information is often not easy to call up.

With SAP HANA Cloud, customers do not have to choose one of the two options, but can benefit from the advantages of both. The interaction of the vector engine and the Knowledge Graph Engine in SAP HANA Cloud enables developers to create context-related, intelligent queries that are far from comparing the search terms.

Let us assume that one user makes the following request: “Find that closest camp (within a radius of ~ 50 km around Frankfurt) by suppliers in Germany that Certified according to ISO 9001 are whose Co₂-tax rates low are and for that No late late in the declaration is.”

It is possible to carry out a query with several models to find the bearings that correspond to the above criteria.

For this purpose, a Sparql table is used in the Knowledge Graph Engine from SAP HANA to determine the suppliers that meet the following conditions: ISO 9001-certified, low CO₂-Tax rates, no reports regarding delays in the payment.

In addition, the function can Sparql_execute in the Knowledge Graph Engine of SAP HANA with a vector -based, semantic filter and spatial restrictions are combined to identify suppliers that are “In a radius of about 50 km around Frankfurt” are located and for the “No late late in the discount ” was reported. This hybrid query uses the vector engine, the Knowledge Graph Engine and the Spatial Engine from SAP HANA Cloud, in order to evaluate suppliers in the area not only in terms of their removal, but also on the basis of their reliability and performance.

After executing these queries, the system delivers the following consignment warehouse as the best hit:

This illustrates the performance of this combination of semantics and structure, which is a central component of SAP HANA Cloud.

Uniform queries: SQL, Sparql and vector search

Developers often have to combine several tools and languages: the database language SQL for relational data, the query language Sparql for RDF (resource description framework, a methodology for the formulation of metadata on the Internet) and separate programming interfaces (APIs) for vector storage.

This complexity is eliminated with SAP HANA Cloud. Users can create a single SQL query that combines relational data, semantic argumentation using Sparql (integrated in SQL) and vector-like search with integrated SQL functions. The result: a uniform in-memory engine, so that no ETL process (extraction, transformation, charging) and no separate infrastructure are required.

This approach not only accelerates the development process, but also enables new types of AI applications that have so far not been practical in environments sealed off.

Designed for generative AI and RAG: Graphrag, Vectrag, Hybridrag

Large voice models (Large Language Models, LLMS) are just as good as the data you can use. That is why Retrieval-Augmented Generation (RAG) has proven to be an important part of generative corporate AI.

SAP HANA Cloud comes up with new functions, for example for training voice models with unstructured text (Vectrag), structured knowledge graphs (Graphrag) or a combination of Vectrag and Graphrag.

SAP HANA Cloud ensures transparency, traceability and performance for all aspects of generative AI and also offers all the advantages of database management. Users receive explainable answers and have complete control over the access, the classification and the compilation of information, which is particularly important in regulated industries.

Effects in practice

Companies from all industries already use the multi-model functions of SAP HANA Cloud to achieve better results:

• Supplier comparison and evaluation with regard to the topics of the environment, society and governance (ESG): Merging structured supplier data with information on the similarity of documents and connections to determine suitable partners
• Compliance monitoring: Integrating and querying guidelines, regulations and audit trails about natural, semantic inputs
• Fraud recognition: Analysis of transaction data, behavioral patterns and well -known fraud patterns in real time
• Research in the area of ​​life sciences: Integration of clinical studies, publications and treatment results using hybrid semantic and structured queries

These are use cases in which the relevance is in formats, systems and connections.

Development processes made easy

SAP HANA Cloud offers developers the following benefits:

• Uniform platform for all data models: Structured, unstructured and semantic data can be combined without having to combine several systems such as a patchwork.
• Integrated support for modern AI workloads: Application cases such as RAG are also possible without external vector memory or pipelines.
• Close integration with SAP and open partner networks: The SAP Business Technology Platform and popular open source tools can be set up and used with minimal effort.
• Focus on innovation instead of infrastructure: It is no longer necessary to manage and maintain separate triplestor databases, search engines or vectord data banks.

Companies can create prototypes faster and benefit from cleaner architecture and less complex operational processes.

New approaches for databases

In companies that focus on AI, data is not only a topic that plays an important role behind the scenes, but also an important driver of innovation. And to promote innovations, you need flexible, intelligent and uniform infrastructure.

SAP HANA Cloud provides building blocks that can be used to create an easily accessible infrastructure for AI apps. This not only facilitates AI workloads, but also accelerates them-with a single platform, the semantics, the detection of similarity and structure together in real time.

A AI needs more than just access to data, and SAP HANA Cloud meets this requirement.

Most important knowledge

• Uniform multi-model approach: Vector, diagram, geo-, text and relational data on a platform
• Intelligent queries: Creation of intelligent queries using SQL, Sparql and Vector searches
• Generative KI: Support from Graphrag, Vectrag and Hybridrag with complete explanability
• Less complexity: No more need for separate vector memory or know -leather graph engines

More information

Philipp Herzig is CTO and Chief Ai Officer and a member of the extended board member of SAP SE.
Stefan Bäuerle is Senior Vice President and Head of SAP BTP/SAP HANA & PERSISTENCY at the SAP.