AlloyDB AI

Automated embeddings generation helps you easily transform operational data, including text, images, and video, into vector embeddings using text embedding or multimodal embedding models. Store embeddings in AlloyDB together with your relational data and query across all data types with a SQL interface.

The ScaNN index uses the same search algorithm as Google Search and is based on 12 years of Google research. It performs advanced semantic search with up to 10x faster index creation, up to 4x faster vector search queries, and up to 10x faster filtered vector search queries than the standard PostgreSQL HNSW index. AlloyDB AI offers additional enhancements such as parallel index build, index auto-maintenance, and enterprise-grade observability for vector indexes.

AlloyDB AI’s ScaNN index is deeply integrated with the PostgreSQL query planner to enable simple yet powerful queries across structured and unstructured data, so you don’t need to deploy or learn a separate vector database. Adaptive filtering ensures that filters, joins, and vector indexes deliver optimal performance when used together.

With AlloyDB AI query engine, you can use natural language in SQL queries to express filtering conditions and ranking criteria. The power of AI models brings reasoning and real-world knowledge to SQL queries, unlocking deep semantic insights from your enterprise data. Sign up for a preview of AlloyDB AI query engine.

Use AlloyDB AI to provide users and agents with accurate responses to natural language questions. With AlloyDB AI natural language you can overcome the ambiguity, flexibility, and security issues typically encountered in natural language interfaces. AlloyDB AI can disambiguate user questions, incorporate data from the schema, sample data, and other sources to increase accuracy, and lock down access to unauthorized data.