AlloyDB AI

AlloyDB AI runs vector queries up to 10x faster compared to standard PostgreSQL. It supports 8-bit quantization for a 3x reduction in index size when enabled, and offers vectors with 4x more dimensions when enabled. Your apps can perform fast similarity searches on complex data types such as text and images, using ANN (approximate nearest neighbor) or KNN (exact k-nearest neighbor) algorithms.

Automated embeddings generation helps you easily transform operational data, including text and images, into vector embeddings using the AI model of your choice. Store embeddings in AlloyDB and query them with a SQL interface.

Access both local models in AlloyDB and remote models hosted in Vertex AI, including custom and pretrained models. You can train and fine-tune models with the data stored in AlloyDB and then deploy them as endpoints on Vertex AI.

Easily build gen AI applications that are more accurate, transparent, and reliable with LangChain integration. Use Document Loaders for loading and storing information from documents, Vector Stores for enabling semantic search, and Chat Messages History for enabling chains to recall previous conversations.

As part of AlloyDB, AlloyDB AI benefits from the best of Google and the best of PostgreSQL. Give your application superior performance and scale, a high availability SLA of 99.99% that's inclusive of maintenance, automatic detection of and recovery from most database failures, and comprehensive security and compliance.