SAP HANA Cloud is SAP’s flagship database delivered as a service from the cloud as part of the Business Technology Platform (BTP). As of today, SAP HANA Cloud is the second most consumed service within BTP and growing at an increasing rate as SAP, customer, and partner engineering teams are building the intelligent data applications that solve today’s most complex business problems.
For customers looking to modernize legacy applications or aligning to a clean core strategy, these types of applications take full advantage of the database's advanced features like multi-tier data storage, advanced multi-model engines, and machine learning. Together these key features enable intelligent solutions to enhance the user experience and effectiveness within the business processes.
The SAP HANA Cloud vector engine is a natural fit to the already existing multi-model suite of engines and provides the data foundation for a complete new set of natural and intuitive capabilities. One example is using the vector engine to augment large language models with business context. In this case, the vector engine accomplishes two purposes. The engine stores relevant business data as vector embeddings. Second, similar and relevant vector embeddings based on the user’s prompt are passed along to a LLM for accurate business answers.
Just as a sidenote, SAP HANA Cloud works alongside the other services within BTP forming the completion solution stack that intelligent data applications need to drive today’s business problems forward.
Without going into much more detail, the fact that SAP HANA Cloud includes a vector engine, is a remarkable innovation. During the past few months, I have had the pleasure to experience the excitement from customers, partners, and even internal engineering teams as they begin to design new intuitive solutions. I know there are many others that share the same excitement. So, I would like to share some of the most common questions I have been asked in order share the news of how the SAP HANA Cloud vector engine powers intelligent data applications.
Q: Why use a vector engine?
A: A vector engine is utilized to empower generative AI applications. By providing an ability to store various data types in the form of vector embeddings, the vector engine facilitates tasks such as Retrieval Augmented Generation (RAG), natural language processing, image recognition, and recommendation systems. These capabilities enable organizations to extract valuable insights from their data, enhance decision-making processes, and develop innovative applications across diverse industries. Furthermore, the integration of the vector engine with SAP HANA Cloud opens up possibilities for seamless data retrieval, augmentation, and analysis, further amplifying its utility and relevance in modern AI-driven workflows.
Q: Why choose a multi-model database over a single purpose database?
A: Opting for a multi-model database provides greater flexibility since it can manage various data types, unlike a single-purpose database that's restricted to one type. Here's why leveraging SAP HANA Cloud with its multi-model data processing capabilities is advantageous:
- Unified Data Management: Store and manage all your data, including vectors, within a single secure platform.
- Simplified Data Interaction: Leverage familiar SQL commands to work with all data types, including vectors.
- Deeper Data Exploration: Combine various data formats (spatial, graph, JSON, etc.) with vector-based queries for richer insights.
- Flexible Development: Integrate vector use cases into your solutions using diverse tools like SAP HANA Cloud clients (Python), hana-ml, and SAP CAP
- Ability to Re-use/apply existing authorizations: With the vector engine integrated into SAP HANA Cloud, customers can reuse existing authorizations, simplifying setup and improving security. This streamlines architecture and operations for smoother workflows and easier maintenance
Q: How can LangChain be used with SAP HANA Cloud to build GenAI solutions?
A: SAP HANA Cloud Vector Engine is accessible from Langchain (both python and java script), opening up a wide array of possibilities for developers to create applications powered by Large Language Models (LLMs). This integration enables seamless synergy between the vector engine and Langchain, allowing developers to leverage the power of both technologies for innovative and powerful applications. For detailed guidance on how to utilize LangChain to create embeddings based on website content and feed them into SAP HANA Cloud, particularly in the context of the RAG scenario, refer to the in-depth blog by our engineering colleague @MartinKolb.
Q: Does the SAP Cloud Application Programming (CAP) model support the use of the SAP HANA Cloud vector engine?
A: Yes, the SAP HANA Cloud Vector Engine is fully supported by CAP, providing developers with a comprehensive toolkit for building scalable and efficient applications. For more details refer to our official reference architecture guidelines for RAG scenarios, and get a practical hands-on experience through our discovery mission - Harnessing Generative AI Capabilities with SAP HANA Cloud Vector Engine
Q: Which embedding model works best with the SAP HANA Cloud vector engine?
A: SAP HANA Cloud offers flexibility by not mandating a specific embedding model. Various data types and use cases may require different approaches. Additionally, the SAP HANA Cloud Vector Engine has a possibility to allow users to bring their own expertise by using custom embedding models they have trained for their specific data and needs.
Q: Which data chunking methods work best with the SAP HANA Cloud vector engine?
A: Though SAP HANA Cloud doesn't offer specific data chunking recommendations at present, it is advisable to divide large documents or inputs into smaller segments before embedding. For textual data, consider the token limit of language and embedding models, whereas for images, lowering resolution is preferable. Your data's attributes should guide your strategy, allowing exploration of chunking techniques like fixed-size or variable-sized methods, considering performance metrics as well.
Q: How does the import of vector data work in SAP HANA Cloud?
A: The import of vector data in SAP HANA Cloud typically involves utilizing SQL commands or supported import tools to upload vector embeddings into the database (for example via SAP HANA Cloud database explorer). However, a prevalent method involves utilizing application interactions, especially for vector storage and querying. For instance, imagine an intelligent application built atop SAP HANA Cloud; this application takes charge of reading and writing vector data, ensuring a seamless flow in and out of the platform. Such dynamic interactions, where the application assumes a central role, exemplify streamlined approaches to housing vector embeddings in SAP HANA Cloud.
Q: How can we transform our standard data, such as tabular or structured / unstructured data, into a vector embedding format?
A: This can be achieved using the embedding functions, a type of machine learning function, to convert categorical data into a format understandable by AI systems. These functions can process various types of information like text, images, or sounds, transforming them into vector embeddings.
Q: Does the data we use to generate the embeddings need to reside solely within the SAP HANA Cloud database itself?
A: No, it's not mandatory. SAP HANA Cloud is open and can connect with many external systems. In addition, data stored in SAP HANA Cloud can be combined with external data. Note that since SAP HANA Cloud is a multi-model database, storing the source data in SAP HANA Cloud opens the opportunity to use the same dataset for other use cases.
Q: How is data security ensured when transmitting vectors to other large language models (LLMs), and are data masking techniques implemented?
A: The same security measures applied to any SAP HANA Cloud instance are also applicable to vector embeddings and tables stored within SAP HANA Cloud, including data masking and anonymization efforts. For more details, please refer to Security in SAP HANA Cloud help document.
Q: Will SAP HANA Cloud also handle the embedding process?
A: Currently, the embedding process needs to be performed outside of SAP HANA Cloud.
To conclude, if you have more questions or if there's anything else on your mind regarding the SAP HANA Cloud Vector Engine, feel free to drop them in the comment section below. Your input is highly valued, and we are here to address any further inquiries you may have. Thank you for taking the time to read this blog, and we look forward to assisting you further on your journey with SAP HANA Cloud.
Next Steps & More Resources:
- Learn more about Vector Engine and its capabilities in SAP HANA Cloud with the official help document
- Tune into watch our recent webinar on Better Business Outcomes with the SAP HANA Cloud Vector Engine
- Try your hands-on with the SAP HANA Cloud Guided Experience
- Gain a better understanding of SAP HANA Cloud’s multi-model engines
- SAP Managed Tags:
