Vanna AI is an open-source Python framework and hosted platform that uses retrieval-augmented generation (RAG) to convert natural language questions into SQL queries against databases like Snowflake, BigQuery, PostgreSQL, and MySQL. It learns from your schema, documentation, and past queries to improve accuracy over time. The hosted cloud tier adds RBAC, audit logs, a vector database, and an embeddable UI. It targets data teams and business users who want self-service analytics without writing SQL. Key features: - RAG-based text-to-SQL generation with schema and history learning - Support for 10+ databases: PostgreSQL, MySQL, Snowflake, BigQuery, SQLite - Multiple LLM backends: OpenAI, Anthropic, Gemini, Ollama - Embeddable UI for web app integration - Role-based access control and audit logs (cloud tier) - Open-source core with Jupyter, Slack, and Flask integration
Open source: free (self-hosted). Explorer: $50/month (20 questions/day, 2-3 users). Team: $500/month (300 questions/day). Enterprise: custom. Annual plans: 20% discount.