SingleStore is updating its unified database to handle natural language queries by converting them into SQL – and it’s building its own AI assistant to boot.
SingleStore provides the real-time transactional and analytic SingleStoreDB database, a distributed, relational SQL database with operational, analytical, and vector data support. It is integrated with Apache Iceberg, Snowflake, BigQuery, Databricks, and Redshift, and CEO Raj Verma positions SingleStoreDB as a foundational layer for AI. SingleStore was taken into majority private equity ownership last month. The database has been undergoing recent developments to add AI/ML functionality, and now even more has been added.
VP of Product Dave Eyler stated: “With these additions, SingleStore becomes for data and AI apps what ChatGPT has become for consumers – an indispensable foundation that turns questions into insights.”
The SingleStoreDB update has three main features:
- AI and ML Functions: Developers can now call machine learning models and large language models (LLMs) directly from SQL, streamlining the process of building intelligent, data-driven apps with familiar syntax.
- Zero Copy Attach: This new functionality makes it easier to experiment on production data by replicating it into lightweight clusters within the same cluster group, enabling safe, agile iteration, especially when paired with SingleStore’s database branching.
- Aura Analyst: An analytics assistant that allows users to conversationally explore insights across all their data. In the future, SingleStore will add the capability to embed Aura Analyst directly into its customers’ apps, turning every application into an AI-native experience and positioning SingleStore as the ChatGPT for data and AI apps.
Aura will provide a natural language-based query interface into SingleStoreDB’s data. It analyzes the query, turns it into a SQL request, sends that to SingleStoreDB data, receives a response, and builds an answer to the input request.
This raises the question of where in your database estate is the appropriate place to have an AI assistant querying your data. We’re seeing a tremendous tussle going on between the various unified data store providers – Databricks, SingleStore, Snowflake, and others – to be the AI model or agent query entry point. Whichever supplier owns that entry point effectively owns the data being queried and becomes the top-level, AI-relevant data store for that customer. Lower-down-the-stack data stores then become mere data feeders to the chosen entry point.
You can view demos of database access with Zero Copy Attach here, running AI tasks directly in SQL with AI Functions here, and Aura Analyst here.
