Add option to persist/archive vector search embeddings

[reeswag]

Can I suggest adding an option to persist/archive embeddings used for vector search, rather than storing them temporarily and clearing them on re-analysis?

I’m working with a very large dataset processing it in bulk and then validating results gradually using the database reload feature. In this workflow, retaining embeddings would allow vector search to remain usable without needing to regenerate them each time.

This would be especially helpful for long-term monitoring projects allowing quick backward searches of large datasets without needing to re-process the data fully!