Datasette Agent alpha brings conversational querying to SQLite databases
Simon Willison released the first alpha of Datasette Agent, a plugin-extensible AI assistant for querying SQLite data in plain language.
Simon Willison has released the first alpha of Datasette Agent, a conversational AI assistant built into his open-source Datasette data exploration tool. The agent lets users ask questions about data stored in SQLite databases in plain language, and can be extended through Datasette's plugin system to add extra tools and capabilities. It runs as a web app in the browser and supports multiple model backends, including Gemma 4 and Qwen 3.5 and 3.6, with the most reliable results on models above 4 billion parameters.
For developers building data-facing applications, the release illustrates a practical pattern for adding natural-language interfaces to SQLite-backed tools without committing to a single model provider. The plugin architecture keeps the core extensible, and multi-model support — spanning open-weights and proprietary APIs alike — gives teams flexibility in how they deploy it. An interactive demo is available, and Willison has published additional details on the Datasette project blog.