Vanna AI

Vanna AI is an open-source Python framework that converts natural-language questions into SQL using retrieval-augmented generation over your schema, documentation, and example queries. It connects to major databases, supports local or hosted LLMs, and is self-hostable, making it well suited to developers building private, accurate text-to-SQL experiences.

Vanna AI is an open-source (MIT-licensed) Python framework for text-to-SQL. Instead of relying purely on a large language model to guess queries, Vanna uses retrieval-augmented generation: you train it on your database schema, documentation, and example question-and-SQL pairs, and it retrieves that context at query time to produce more accurate SQL. It connects to a wide range of databases, including Snowflake, BigQuery, Postgres, and Athena, and can pair with hosted models or local LLMs via tools like Ollama. The framework ships with a production-ready web interface and can be embedded into applications, notebooks, or internal tools. Because it is code-first and self-hostable, teams can keep their data and query context in their own environment, which appeals to organizations with privacy or compliance requirements. A significant rewrite (referred to as Vanna 2.0) in late 2025 moved the project from a SQL-generation library toward a more complete agent framework, adding capabilities such as row-level security and support for current-generation models. Vanna is aimed at data teams, analytics engineers, and developers who want to build natural-language query experiences over their own databases without sending schema and data to a closed vendor. Its open-source core is free; the project and associated hosted offerings sit alongside a growing category of AI SQL assistants.