4.6 Performing a Vector Search With Retrieval-Augmented Generation

4.6.1 Running Retrieval-Augmented Generation

4.6.2 Using Your Own Embeddings With Retrieval-Augmented Generation

When you enter a query, HeatWave GenAI performs a vector-based similarity search to retrieve similar content from the vector store and embedding tables available in the DB system. It provides the retrieved content as context to the LLM. This helps the LLM to produce more relevant and accurate results for your query. This process is called as retrieval-augmented generation (RAG).

As of MySQL 9.2.1, you can use both inbuilt vector store tables and tables containing your own vector embeddings for running RAG with vector search.

The topics in this section describe how to perform RAG with vector search.