详细内容
Vanna 2.0
DATA-FIRST AGENTS
User-scoped execution, rich UI components, and production features built-in—not bolted on.
Vanna 2.0 puts user identity at the center. Every request carries permissions, every tool checks access, and every response is optimized for both LLMs (token efficiency) and humans (rich displays). From prototype to production without the rewrite.
Get Started Free Install Open Source
Context-Aware
User & workspace permissions
Rich Updates
Structured UI components
Dual Output
Optimized for LLM & display
How It Works
Simple, secure, and user-aware from the first request
🌐
Your App
<vanna-chat>
Drop-in webcomponent
🤖
User-Aware Agent
- ✓ Knows who's asking
- ✓ Checks permissions
- ✓ Rate limits
🛠️
Your Tools
- • Database queries
- • API calls
- • File operations
✨
Rich UI
- • Interactive charts
- • Data tables
- • Live components
Interactive results streamed back to your users
Production-ready from line one
Built for production, not just demos
User-scoped execution, stateful components, and intelligent outputs. No rewrites required when you scale.
🔐
User identity drives execution
Every request carries user ID, permissions, and workspace context. Tools check permissions automatically, conversations stay isolated, and quota limits enforce themselves.
🎨
Components update, not just append
Status cards transition from "running" to "success". Progress bars fill from 0% to 100%. Log viewers add entries. Unlike chat that appends text, Vanna components maintain state.
⚡
Dual outputs save tokens
Query 1,247 database rows? LLM sees "Query returned 1,247 rows" (5 tokens). User sees a full sortable data table (0 tokens to LLM). Both get exactly what they need.
🗄️
Built for data analysis
Execute SQL queries, return interactive data tables with pagination, sort, filter, and CSV export. No more dumping massive datasets into chat context.
📊
Interactive artifacts
Agents generate Plotly visualizations, HTML dashboards, or custom widgets that render externally—like Claude's Artifacts, but for your product.
🛡️
Production-ready features
Per-user quota management, usage tracking, conversation persistence, and automatic permission enforcement. Built for real products, not just demos.
"
Why use AI to chat with your database in the first place? To start, it allows you to spend less time writing SQL and more time generating insights on your data. With Vanna, you can empower users across your organization—regardless of their SQL expertise—to interact with your database using natural language, making data more accessible to everyone.
Brian Spindolini
Senior Product Manager, Azure SQL Database
Read more on the Microsoft blog
Choose Your Stack
Vanna works with any LLM provider and database. Select your options to get started with the open source framework.
🧠 LLM Provider🤖 Anthropic✨ OpenAI💎 Gemini🦙 Ollama
🗄️ Database📄 SQLite🐘 PostgreSQL🐬 MySQL❄️ Snowflake📊 BigQuery 📚 Other Databases - See Docs →
☁️ Optional Hosted
Vanna provides optional hosted services
- ✓Observability: Real-time metrics & tracing
- ✓Agent Memory: Hosted vector DB
- ✓File System: Secure file storage
- ✓Audit Logs: Activity tracking
- ✓Data Retention: Long-term storage
🧠 LLM Provider
Choose your language model
🤖
Anthropic
✨
OpenAI
💎
Gemini
🦙
Ollama
🗄️ Database
Choose your data source
📄
SQLite
🐘
PostgreSQL
🐬
MySQL
❄️
Snowflake
📊
BigQuery
📚\
\
Other →
☁️ Optional Hosted
Vanna provides optional hosted services
- ✓Observability: Real-time metrics & distributed tracing
- ✓Agent Memory: Hosted vector database for learning
- ✓File System: Secure file storage and retrieval
- ✓Audit Logs: Comprehensive activity tracking & compliance
- ✓Data Retention: Long-term storage for conversations & logs
Use Vanna OSS with These Options →
Get copy-paste ready configuration code