What is Vanna 2.0?
Vanna 2.0 is a production-grade Python framework for building conversational AI agents that are user-aware, permission-conscious, and capable of rich interactive experiences.
Each Vanna Agent connects an LLM to your data and tools through a permission-aware registry, returning rich interactive components to the user.
Interactive Feature Overview
Figure: High-level architecture of a Vanna Agent — connecting users, tools, LLMs, and databases in a composable pipeline.
🔄 Migration and Evolution
What’s New in v2?
If you’re coming from Vanna 0.x (the SQL generation framework), v2 is a complete reimagining:
| Vanna 0.x | Vanna 2.0 |
|---|---|
| SQL generation via RAG | General-purpose agent framework |
| Single-shot Q&A | Multi-step tool execution |
| Anonymous requests | User-aware from the first token |
| Text responses only | Rich UI components (tables, charts, status cards) |
| Fixed architecture | Composable providers (LLM, storage, observability) |
| Training-based | Tool-based with optional RAG |
See the Migration Guide for detailed transition steps.
🚀 Getting Started
Quickstart with Sample Data
Use the sample Chinook SQLite database
Quickstart with Your Own Data
Connect to your database in Python
Migrate from 0.x
Moving from Vanna 0.x to v2
💬 Community and Support
- GitHub Repository
- Issues – Report bugs or request features
- Discussions – Ask questions and share examples