Vectorize your Data : SAP HANA Cloud's Vector Engine for Unified Data Excellence

[shabana]

Introduction:

In the ever changing field of database development, where innovations are a perennial topic, anticipation is building for the release of SAP HANA Cloud's Vector Engine scheduled for March 2024. As technology leaders, database developers are actively seeking Artificial Intelligence (AI) and Generative AI (GenAI) services to navigate the evolving landscape, SAP, recognizing this demand, is all geared up to unveil the Vector Engine. This is an important advancement and addition to the existing multi-model engines in SAP HANA Cloud.

SAP HANA Cloud Vector Engine: Retrieval Augmented Generation Use Case Overview

The upcoming launch of SAP HANA Cloud's Vector Engine not only signifies a leap forward in addressing the challenges associated with solving certain GenAI scenarios but also marks a breakthrough in overcoming data silos. The challenge also lies in finding the delicate equilibrium between efficient data management and handling the complexities of application logic. The Vector Engine is positioned to offer an innovative solution to this dilemma. Amid discussions around data silos and processing latency, SAP HANA Cloud is preparing to introduce its Vector Engine—a new, multi-model engine that empowers users to store and query vector embeddings seamlessly, treating them like any other data types.

Storing vector embeddings within the same database is a strategic move that aligns seamlessly with SAP's commitment to providing a unified platform. This integration eliminates the hurdles posed by data silos, offering a holistic approach to data management. In SAP HANA Cloud, the storage of vector embeddings is seamlessly integrated into the platform's existing structure, allowing users to store them in a designated table. Developers can perform SQL-like queries effortlessly. This means you can execute joins, apply filters, and perform selects by combining vector embeddings with various data types, including transactional, spatial, graph, and JSON data, all within the same SQL environment. The Vector Engine ensures a user-friendly experience, eliminating the need for extensive learning or the adoption of new querying methodologies. Essentially, working with vector embeddings in SAP HANA Cloud is as straightforward as crafting queries in a standard SQL database, offering familiarity and ease of use for developers.

Planned Features:

The Vector Engine in SAP HANA Cloud will:

1. Facilitate storage of vector embeddings by introducing:

• A new data type named REAL_VECTOR
• A vector constructor TO_REAL_VECTOR, to create vector from strings

2. Facilitate similarity search queries and analysis by introducing:

• Two new distance calculating similarity search functions, L2Distance() and cosine_similarity(), to enhance the platform's capability to compute vector similarity.

These features in turn will facilitate:

• Storage and querying of vector embeddings in SAP HANA Cloud through SQL

• In-memory similarity searches to support retrieval-augmented generation (RAG) patterns

• Power to combine business data with graph, spatial, document, and vector data all on a single platform

• Storage and retrieval of contextual information for GenAI and intelligent data applications as vector embeddings

Vector Embeddings & Practical Use Cases:

Vector embeddings are mathematical representations used to encode objects into multi-dimensional vector space. These embeddings capture the relationships and similarities between objects. SAP HANA Cloud Vector Engine will facilitate the storage and analysis of complex and unstructured vector data(embeddings) into a format that can be seamlessly processed, compared, and utilized in building various intelligent data applications and adding more context in case of GenAI scenarios.

Among the most compelling use cases to explore with the vector engine are:

• Semantic Search and Retrieval: Leverage vector embeddings to enhance semantic search capabilities, enabling users to find relevant information quickly.

• Contextual Analysis: Use vector embeddings for contextual analysis, allowing for a deeper understanding of relationships between different data points within the database.

• Intelligent Data Applications: Utilize vector embeddings to build intelligent data applications, unlocking insights and facilitating more informed decision-making.

• Enhanced Recommendations: Leverage vector embeddings to improve recommendation systems, providing users with more accurate and personalized suggestions.

• Optimized Large Language Models (LLMs): Enhance the output of LLMs by utilizing vector embeddings to optimize and add context to the generated content.

SAP HANA Cloud is the Center of our AI Strategy:

As the Vector Engine becomes an integral component of SAP HANA Cloud's multi-model area, it aligns seamlessly with SAP's GenAI strategy. Our vision is to facilitate Gen AI in business context and to achieve this we are enhancing the SAP Business Technology Platform (SAP BTP) with Gen AI capabilities. SAP BTP will provide centralized access to SaaS-based large language models (LLMs) from multiple vendors and host LLMs from open-source models or third parties. Additionally, we plan to develop foundation models integrating our own unique structured data and business process knowledge.

An intriguing case worth exploring involves the synergy of vector embeddings within Large Language Models (LLM) and GenAI scenarios, especially when utilized alongside the Retrieval Augmented Generation (RAG) pattern. These embeddings optimize LLM output by providing contextual information through RAG, acting as a bridge between the query and the vast pool of knowledge stored in SAP HANA Cloud. When an end user queries, the system uses vector embeddings to offer context to the LLM, facilitating detailed and accurate answers. This dynamic interaction enhances content quality, ensuring it is contextually relevant to specific organizational nuances. Consequently, the integration of vector embeddings in LLM scenarios not only enhances information retrieval precision but also boosts the adaptability and intelligence of language models, resulting in sophisticated and context-aware language processing.

In addition, SAP maintains a commitment to openness. Aligning with our Gen AI strategy, if customers utilize their own Large Language Models (LLMs) and/or Retrieval Augmented Generation (RAG) technology, SAP HANA Cloud remains the preferred choice. The platform's compatibility and openness ensure that organizations can seamlessly integrate their existing technologies, aligning with SAP's dedication to providing flexible solutions tailored to customer needs.

Here's a video providing a comprehensive look at the SAP HANA Cloud Vector Engine, highlighting its added value and featuring a brief demonstration of its functionality.

 

Conclusion:

The inclusion of the Vector Engine as part of SAP HANA Cloud provides intuition to new and existing apps without an additional data silo. SAP HANA Cloud envisions a future where data becomes smarter, faster, and more efficient. As we eagerly anticipate the launch of SAP HANA Cloud's Vector Engine in March 2024, we invite you to stay tuned for future updates. The journey doesn't end here, there is much more to explore and unveil as we approach the release date. We'll be sharing insights, additional use cases, and exciting features that make the Vector Engine an important part of the innovative and intuitive solutions.

References:

1. SAP HANA Cloud’s Vector Engine Announcement
2. Free SAP HANA Cloud Guided Experience: Building a Data Foundation for Intelligent Data Applications
3. Recap SAP HANA Cloud @ SAP TechEd