Document embeddings

[mitchelljkotler]

[eyeseast]

Choices a user might make around embeddings:

• What model (and settings) do I use?
• What am I embedding? A whole document, a page, a chunk of text?
• Am I embedding PDFs directly, or images, or extracted text?

Our considerations:

• Where do we store embeddings?
• This feels like a premium feature.
• Is this in DocumentCloud core, or is it a separate application?