GenAI Reference Solution Architecture on SAP Business Technology Platform

Estimated read time 11 min read

Introduction

In a remarkably short span, GenAI has taken the world by storm. With the help of Large Language Models (LLMs) we can tap boundless opportunities for everyone. Enterprises globally can harness the power of GenAI to innovate and streamline their processes like never before.

SAP is at the forefront of this revolution, making substantial investments in GenAI. With its robust, enterprise-ready Business Technology Platform (BTP), it enables the creation of GenAI solutions that seamlessly integrate with both SAP and non-SAP cloud services and customer’s S/4 HANA ERP system.

In this blog, I present a solution architecture for a side-by-side extension that not only integrates seamlessly with S/4 HANA but also leverages the cutting-edge Generative AI services on the Business Technology Platform. Discover how this powerful combination can transform your enterprise, driving efficiency and unlocking new possibilities for growth and innovation.

I will structure this blog in 2 sections.

Architectural Blueprint: A detailed architecture diagram illustrating the design.Component Breakdown: In-depth explanations of each component, highlighting it’s importance and benefits in the architecture.

Let’s get started. 

 

Architectural Blueprint

The following diagram presents an architectural blueprint for GenAI solution or S/4 HANA extension, developed and deployed on BTP. This architecture ensures seamless integration with on-premise S/4 HANA through connectivity service and leverages hyperscaler LLMs via the Generative AI Hub to unlock AI capabilities. 

Architecture Diagram

 

Let’s deep dive into each component.

 

Component Breakdown

In this section, we’ll take an in-depth look at each component. Understanding these components and their responsibility will provide a comprehensive view of how they collectively contribute to the solution.

Business Technology Platform (BTP):

This is the platform for creating an Intelligent Enterprise. It is an integrated offering comprising of database, application development, integration and intelligent technologies like GenAI. As you can see in the architecture diagram, you can create and integrate your solution completely leveraging cloud services on BTP. As you can see some of the many services leveraged in the architecture as listed below

HANA Cloud databaseGenerative AI HubCloud Foundry RuntimeSAP Build Work ZoneConnectivity service

We will be going in details of each one of it in the following sections.

HANA Cloud Database:

SAP HANA Cloud is SAP’s database-as-a-service offering, giving you the power and performance of SAP HANA in the cloud. The SAP HANA Cloud allows you to consume the SAP HANA database from applications running on SAP Business Technology Platform.

SAP HANA Cloud vector engine provides capability to create relevant and responsible GenAI solutions. Recent advances in GenAI and Large Language Models has generated importance for a reliable vector databases. Similarity search complements relational database and full-text search systems. Embedding functions map natural language text to high-dimensional vectors. Since LLM data can lead to hallucinations, vector databases can complement LLM responses with additional relevant text through Retrieval Augmented Generation (RAG). Leveraging vector database like SAP HANA Cloud can ensure response quality from LLM.

Generative AI Hub:

Generative AI Hub provides you access to diverse set of LLMs which can be integrated with your solution on SAP BTP in a cost effective way. It gives access to SAP foundational models and open source models. In the architecture diagram, I have considered Azure OpenAI LLMs which are described in brief below

gpt-4o-turbo: Useful for generating relevant response for applications based on prompt and content. Assisting language translations and summarization.text-embeddings-ada-002: For creating embeddings for the content or summary. Used for clustering similar text together. The embeddings are stored in vector database which can be referred at later point of time.

Cloud Foundry Runtime:

Cloud foundry run time helps you develop and run cloud native solutions. It is built on open standards where developers can choose language of their choice. For example – Node.js, JAVA or Python.

You can create applications leveraging CAP (Cloud Application Programming) model and access underlying custom tables in HANA cloud database using declarative programming paradigm with CDS (Core Data Services). I have showcased a full stack CAP application deployed on cloud foundry runtime.

It also provides Python build packs for creating AI driven microservices. In the architecture diagram, you can see the Python microservice is integrating with LLMs using python libraries (generative-ai-hub-sdk & ai-core-sdk). Using opensource library like Langchain, you can seamlessly integrate with vector engine in SAP HANA Cloud. These python libraries are available to install from python packaging index (PyPi).

SAP Business Application Studio provides support for development for all the above mentioned languages to ease the development process.

SAP Build Work Zone: 

SAP Build Work Zone enables organizations to establish a centralized point of access to SA custom-built, and third party applications and extensions. As shown in the Architecture diagram, the free style UI5 CAP application is integrated in the SAP Build work zone, from where it can be accessed based on roles configured for the user.

Connectivity Service:

SAP BTP Connectivity service enables cloud applications deployed on BTP to securely access remote services on the Internet or on-premise. It allows configuration of application connections via destinations and facilitates connections to on-premise systems using the Cloud Connector. This component also establishes a secure tunnel with your on-premise network so your applications can communicate with on premise systems securely with trust.

As shown in the architecture diagram, CAP service integrates with S/4 HANA APIs (On Premise) securely through cloud connector which is configured on the customer premises. CAP service integrates with other cloud applications via destinations configured in the connectivity service.

 

Conclusion

To conclude, the integration of SAP Business Technology Platform (BTP) with advanced Generative AI (GenAI) and Large Language Models (LLMs) offer countless opportunities for enterprises to scale efficiently. By leveraging the robust capabilities of vector databases and Retrieval Augmented Generation (RAG), businesses can significantly enhance the accuracy and quality of AI-driven solutions.

The architectural blueprint provided in this blog demonstrates how seamless integration with on-premise S/4 HANA systems and the utilization of hyperscaler LLMs through the Generative AI Hub can be achieved leveraging cloud services to create side by side extension on BTP thus following the “Clean Core” approach.

Note: The content in this blog is based on my experience working on SAP BTP and creating AI solutions on BTP.

To get more updates about this topic, please follow the below pages

SAP Business Technology Platform Generative AI Hubgenerative-ai-hub-sdkLearning Journeys

Feel free to “like“, “Share“, “Add a Comment” and to get more updates about my next blogs follow me – avinash.vaidya@sap.com

 

 

​ IntroductionIn a remarkably short span, GenAI has taken the world by storm. With the help of Large Language Models (LLMs) we can tap boundless opportunities for everyone. Enterprises globally can harness the power of GenAI to innovate and streamline their processes like never before.SAP is at the forefront of this revolution, making substantial investments in GenAI. With its robust, enterprise-ready Business Technology Platform (BTP), it enables the creation of GenAI solutions that seamlessly integrate with both SAP and non-SAP cloud services and customer’s S/4 HANA ERP system.In this blog, I present a solution architecture for a side-by-side extension that not only integrates seamlessly with S/4 HANA but also leverages the cutting-edge Generative AI services on the Business Technology Platform. Discover how this powerful combination can transform your enterprise, driving efficiency and unlocking new possibilities for growth and innovation.I will structure this blog in 2 sections.Architectural Blueprint: A detailed architecture diagram illustrating the design.Component Breakdown: In-depth explanations of each component, highlighting it’s importance and benefits in the architecture.Let’s get started.  Architectural BlueprintThe following diagram presents an architectural blueprint for GenAI solution or S/4 HANA extension, developed and deployed on BTP. This architecture ensures seamless integration with on-premise S/4 HANA through connectivity service and leverages hyperscaler LLMs via the Generative AI Hub to unlock AI capabilities. Architecture Diagram Let’s deep dive into each component. Component BreakdownIn this section, we’ll take an in-depth look at each component. Understanding these components and their responsibility will provide a comprehensive view of how they collectively contribute to the solution.Business Technology Platform (BTP):This is the platform for creating an Intelligent Enterprise. It is an integrated offering comprising of database, application development, integration and intelligent technologies like GenAI. As you can see in the architecture diagram, you can create and integrate your solution completely leveraging cloud services on BTP. As you can see some of the many services leveraged in the architecture as listed belowHANA Cloud databaseGenerative AI HubCloud Foundry RuntimeSAP Build Work ZoneConnectivity serviceWe will be going in details of each one of it in the following sections.HANA Cloud Database:SAP HANA Cloud is SAP’s database-as-a-service offering, giving you the power and performance of SAP HANA in the cloud. The SAP HANA Cloud allows you to consume the SAP HANA database from applications running on SAP Business Technology Platform.SAP HANA Cloud vector engine provides capability to create relevant and responsible GenAI solutions. Recent advances in GenAI and Large Language Models has generated importance for a reliable vector databases. Similarity search complements relational database and full-text search systems. Embedding functions map natural language text to high-dimensional vectors. Since LLM data can lead to hallucinations, vector databases can complement LLM responses with additional relevant text through Retrieval Augmented Generation (RAG). Leveraging vector database like SAP HANA Cloud can ensure response quality from LLM.Generative AI Hub:Generative AI Hub provides you access to diverse set of LLMs which can be integrated with your solution on SAP BTP in a cost effective way. It gives access to SAP foundational models and open source models. In the architecture diagram, I have considered Azure OpenAI LLMs which are described in brief belowgpt-4o-turbo: Useful for generating relevant response for applications based on prompt and content. Assisting language translations and summarization.text-embeddings-ada-002: For creating embeddings for the content or summary. Used for clustering similar text together. The embeddings are stored in vector database which can be referred at later point of time.Cloud Foundry Runtime:Cloud foundry run time helps you develop and run cloud native solutions. It is built on open standards where developers can choose language of their choice. For example – Node.js, JAVA or Python.You can create applications leveraging CAP (Cloud Application Programming) model and access underlying custom tables in HANA cloud database using declarative programming paradigm with CDS (Core Data Services). I have showcased a full stack CAP application deployed on cloud foundry runtime.It also provides Python build packs for creating AI driven microservices. In the architecture diagram, you can see the Python microservice is integrating with LLMs using python libraries (generative-ai-hub-sdk & ai-core-sdk). Using opensource library like Langchain, you can seamlessly integrate with vector engine in SAP HANA Cloud. These python libraries are available to install from python packaging index (PyPi).SAP Business Application Studio provides support for development for all the above mentioned languages to ease the development process.SAP Build Work Zone: SAP Build Work Zone enables organizations to establish a centralized point of access to SA custom-built, and third party applications and extensions. As shown in the Architecture diagram, the free style UI5 CAP application is integrated in the SAP Build work zone, from where it can be accessed based on roles configured for the user.Connectivity Service:SAP BTP Connectivity service enables cloud applications deployed on BTP to securely access remote services on the Internet or on-premise. It allows configuration of application connections via destinations and facilitates connections to on-premise systems using the Cloud Connector. This component also establishes a secure tunnel with your on-premise network so your applications can communicate with on premise systems securely with trust.As shown in the architecture diagram, CAP service integrates with S/4 HANA APIs (On Premise) securely through cloud connector which is configured on the customer premises. CAP service integrates with other cloud applications via destinations configured in the connectivity service. ConclusionTo conclude, the integration of SAP Business Technology Platform (BTP) with advanced Generative AI (GenAI) and Large Language Models (LLMs) offer countless opportunities for enterprises to scale efficiently. By leveraging the robust capabilities of vector databases and Retrieval Augmented Generation (RAG), businesses can significantly enhance the accuracy and quality of AI-driven solutions.The architectural blueprint provided in this blog demonstrates how seamless integration with on-premise S/4 HANA systems and the utilization of hyperscaler LLMs through the Generative AI Hub can be achieved leveraging cloud services to create side by side extension on BTP thus following the “Clean Core” approach.Note: The content in this blog is based on my experience working on SAP BTP and creating AI solutions on BTP.To get more updates about this topic, please follow the below pagesSAP Business Technology Platform Generative AI Hubgenerative-ai-hub-sdkLearning JourneysFeel free to “like“, “Share“, “Add a Comment” and to get more updates about my next blogs follow me – avinash.vaidya@sap.com    Read More Technology Blogs by SAP articles 

#SAP

#SAPTechnologyblog

You May Also Like

More From Author

+ There are no comments

Add yours