IBM was looking for a powerful search vector database solution to store text, image, and video embeddings for its clients. With the Elasticsearch vector database, not only are all those data formats supported, but there is also a catalog of integrations to bring data from various enterprise data sources, including third-party. Retrieval across disparate data types can also leverage secure and native hybrid search for the most relevant combination of text, vectors, and geospatial data results with filtering, aggregations, and document-level security.
Developers can now implement vector search and semantic search, including k-nearest neighbors (kNN) and approximate nearest neighbor (ANN) search, with flexible multi-cloud model management provided for popular natural language processing (NLP) models and Elastic open inference API. Elasticsearch also includes ELSER, an out-of-the-box sparse encoder model for semantic search, and access to reranking models to improve search results. These capabilities can be used seamlessly with IBM foundation models (like Slate and Granite), available in IBM watsonx.ai.
Elasticsearch is the world’s most downloaded vector database, and our team is constantly investing in its capabilities to take its performance and scale further. More recently, optimizations and innovations like adding scalar quantization and other performance enhancements resulted in up to 8x to 32x efficiency gains for developers building AI-enabled apps.
Leave a Reply