Is it possible to use open source embeddings for the vector similarity search?

[dustyatx]

The default tantivy tokenizers seem like basic NLP tokenization (unless I'm missing something, I'm no expert). I've been using open source embeddings models with other vector databases, is that possible with CozoDB? If so is there any advantage or disadvantage to using them?

I've been mainly using:
e5-large-v2

Also any chance someone has a basic walk through on how to load up tokens/embeddings? Datalog and this flavor of it is a bit of brain bender (I'm slowly getting it).