Support embeddings workflows

Support workflows involving embeddings and vector databases. Need to discuss:

- whether the storage of embeddings should be left to the developer
- if not, ephemeral vs persistent embeddings use cases
- most expensive "tinkering" use cases that require vector databases, that could be made low/zero cost for devs
- cases where multiple apps are computing embeddings on the same data
- costs of switching between embedding models (recomputing embeddings)