WinBuzzer News Google Unveils AlloyDB AI for Generative AI Applications Google AlloyDB AI: A new feature that lets developers create generative AI applications using their operational data. By Luke Jones - August 30, 2023 5:36 pm CEST FacebookTwitterPinterestWhatsApp Google has introduced AlloyDB AI, a new feature integrated into its AlloyDB for PostgreSQL, designed to assist developers in creating generative AI applications using their operational data. This update allows developers to combine the capabilities of large language models (LLMs) with real-time operational data, offering comprehensive support for vector embeddings. AlloyDB AI facilitates the transformation of data into vector embeddings using a straightforward SQL function, and it can execute vector queries up to ten times faster than the standard PostgreSQL. Vector Embeddings: Bridging Data and LLMs The official announcement on Google Cloud Blog by Andi Gutmans, GM & VP of Engineering at Google Cloud Databases, highlighted the significance of vector embeddings in bridging the gap between data and LLMs. These numeric representations of data are essential for understanding the meaning of the underlying data. They play a pivotal role in Retrieval Augmented Generation (RAG) workflows, which use embeddings to find, filter, and represent relevant data to augment LLM prompts. This capability can power experiences like real-time product recommendations, allowing users to search for the most pertinent items. Enhancements Over Standard PostgreSQL While PostgreSQL already offers basic vector support, AlloyDB AI enhances this by streamlining the development experience and boosting performance to cater to a broader range of workloads. This comprehensive solution for working with vector embeddings simplifies the process of building generative AI experiences. Developers can create and query embeddings to find relevant data with just a few lines of SQL, eliminating the need for a specialized data stack or data transfers. Integration with AI Ecosystem AlloyDB AI introduces several capabilities to help developers integrate their real-time data into generative AI applications. These include easy embeddings generation, enhanced vector support, and integrations with the AI ecosystem, such as Vertex AI Extensions and LangChain. Gutmans mentioned in the official announcement, “AlloyDB AI introduces a simple PostgreSQL function to generate embeddings on your data. With a single line of SQL, you can access Google's embeddings models, including both local models for low-latency, in-database embeddings generation, and richer remote models in Vertex AI.” AlloyDB Omni: AI Apps Everywhere AlloyDB AI emphasizes portability and flexibility. With AlloyDB Omni, users can leverage this technology to construct enterprise-grade, AI-enabled applications in various environments, including on-premises, at the edge, across multiple clouds, or even on developer laptops. Many customers already trust AlloyDB for their critical applications. The official announcement highlighted that the Chicago Mercantile Exchange (CME) Group is considering AlloyDB for their most demanding enterprise workloads and is in the process of migrating several databases from Oracle to AlloyDB. For those transitioning from Oracle to AlloyDB, Google has also announced Duet AI in the Database Migration Service, which offers AI-assisted code conversion to automate the conversion of Oracle database code.