What is Vanna.AI?
Vanna AI is a personalized AI-based SQL agent that helps users retrieve actionable insights from their databases by asking questions. Its skills are tied to the training data given to it, meaning that the more data provided, the better its accuracy for handling large and complex datasets.
Features
-
Open-Source Integrations
Users can run Vanna.AI on their own infrastructure and use various frontend integrations like Jupyter Notebook, web apps, Slackbot, and Streamlit apps.
-
Accurate Text-to-SQL Generation
Vanna.AI is excellent at converting natural language questions into precise SQL queries, even for complex datasets. The tool’s accuracy improves with more training data, making it ideal for large and complex database structures.
-
Vector Store Management
The capacity of Vanna.ai to independently manage vector storage is one of its best features. Vanna.ai manages these vectors without the need for manual interaction because they are kept in a vector database.
-
Debugging and Logs
The platform has strong debugging capabilities. Server logs are available for troubleshooting, providing effective operations.
How to Use Vanna.AI?
- Training the RAG Model (vn.train): This step involves providing the model with your data definition languages (DDL) and SQL documentation. The model converts this information into numerical embeddings and saves them in a vector database.
- Querying the Database (vn.ask): Once the model is trained, you can ask queries in natural language. The system converts these questions into vectors, retrieves relevant embeddings from the vector store, and creates SQL queries to execute on your database.
Pros & Cons
Pros:
- Connects to multiple databases.
- Provides continuous performance improvement over time.
- High accuracy mode on complex datasets.
- Private and secure, providing the safety of data.
Cons:
- Limitations in handling extremely complex queries.
- Training data dependence.
- Needs tuning.
- Requires setup and configuration.
Pricing & Plan
Free Plan:
The free plan offers rate-limited use of the large language model for SQL query generation and free use of hosted metadata storage.
Paid Plan:
A premium subscription service costs $15 monthly for production use with no rate limits on SQL query generation using the latest large language model and available service level agreements.
FAQs
Can I use Vanna.AI without an API Key?
Yes, the Vanna.AI API key is only necessary if you're using Vanna-hosted components (i.e. LLM or vector store).
Can I use Vanna.AI with my SQL database?
Yes, Vanna.AI works with any SQL database that you can connect with Python.
How can I see my training data in Vanna.AI? How can I edit it?
Use vn.get_training_data() syntax to see the data frame of your all training data in Vanna.AI.
Summary
I used the Vanna.AI tool for creating SQL queries and realized that there is no need to rely on building SQL queries on other tools which is also very time-consuming. Vanna.AI also offers a Slack integration, and I am considering it a Discord integration if a relevant use case emerges.
Overall, Vanna.ai simplifies database interactions and makes it a valuable tool for various applications.
