Vanna AI
Zilliz Cloud and Vanna combine vector search and RAG for improved SQL generation.
Use this integration for FreeAbout Vanna
Vanna is an open-source Python RAG (Retrieval-Augmented Generation) framework designed for SQL generation and related tasks. It operates in two main steps: training a RAG model on your data, and then generating SQL queries based on user questions. Vanna is built with a focus on security, ensuring that database contents are not sent to the LLM unless specifically enabled.
Vanna offers high accuracy on complex datasets, with its capabilities directly tied to the training data provided. As users interact with Vanna, the model continuously improves through the augmentation of training data. The open-source nature of Vanna allows users to run it on their own infrastructure, providing flexibility and control.
Why Zilliz Cloud and Vanna
Combining Zilliz Cloud and Vanna creates a powerful solution for knowledge engineering and SQL generation. Zilliz Cloud, based on the Milvus vector database, provides efficient vector storage and similarity search capabilities. This complements Vanna's RAG framework, enhancing its ability to retrieve relevant information for SQL generation.
This integration allows developers to build more accurate and efficient SQL generation systems. By leveraging Zilliz Cloud's vector search capabilities, Vanna can more effectively find and utilize relevant training data, leading to improved SQL query generation. This is particularly beneficial for large and complex datasets, where accurate information retrieval is crucial for generating correct SQL queries.
How Zilliz Cloud and Vanna works
The integration of Zilliz Cloud and Vanna works by enhancing Vanna's RAG process with Zilliz Cloud's vector search capabilities. First, relevant data such as database schemas, documentation, and previous queries are embedded and stored as vectors in Zilliz Cloud. When a user asks a question, Zilliz Cloud performs a similarity search to retrieve the most relevant information.
Vanna then uses this retrieved information to generate SQL queries. The vector search capabilities of Zilliz Cloud help ensure that the most relevant context is provided to Vanna's model, improving the accuracy of the generated SQL. As users interact with the system, new queries and feedback can be embedded and stored in Zilliz Cloud, continuously improving the model's performance over time.
Learn
The best way to start is with a hands-on tutorial. This tutorial will walk you through how to build a knowledge engineering application with Vanna & Zilliz Cloud.
Tutorial: Write SQL with Vanna and Milvus
And here are a few more resources:
- Vanna Official Website
- Vanna GitHub Repository