SingleStore: SQL, JSON, full text, and vector in a single platform
Date:
By: Daniel Howard
Classification: InBrief
SingleStore is a translytical database that enables organisations to manage both transactions and analytics in a single platform, minimising data movement and reducing latency. Moreover, it is a fully-featured multi-model database, capable of supporting SQL, JSON, full text, and vector workloads. This removes the need to build and manage a complex, expensive, and often fragile architecture of more specialist databases.
The product’s support for vector data is particularly notable. SingleStore incorporated full vector functionality into its core database in 2017 – which is to say that it is an established capability, not a recent addition – and provides key vector features that can be used to train a bespoke LLM (Large Language Model) or support a RAG (Retrieval Augmented Generation) architecture; in other words, it enables you to effectively leverage generative AI. For instance, several of its customers have found success combining its vector and full text search capabilities to implement hybrid search. This is all supported by a range of integrations with AI products, like LangChain and Hugging Face.