Some experiments with Db2 vector capabilities that demonstrate the power of SQL, but also show that careful selection of embedding models is needed
Here is my howto on searching IBM Documentation, including the Db2 docs. It's a follow-up to a blog post from 2021.
I asked IBM Bob as my development partner to explain me the Db2 access plan shown in my previous blog post.
Similarity search for vector data often is approximate nearest neighbor (ANN) search. Db2 introduces vector indexes to speed up such queries.
Db2 now supports integration of external models. This allows computation of vector embeddings or text generation directly from the SQL context.
Db2 Genius Hub supports several authentication methods for secure access. I show how to set up authentication by using the repository and database roles.
I looked into how to research ETF properties using the Db2 vector functionality. Will it benefit my (early) retirement?
