RAG and PG ts vector

[celalp]

I am using postgres for the knowledgebase for 3 reasons

1. I'm old
2. support for full text (keyword search)
3. support for embeddings and offloading similarity searches to the db

I have never seen an instance where one have used both semantic similarity and some sort of token frequency (see TF/IDF) I think it's a good idea to use both in semantic search but I am willing to be convinced that this is not worthwhile.