Automated extraction of structure + metadata from a URL

For the [i3 data backend](https://github.com/the-interlace/notes/blob/main/i3_data_backend.md):
How much structured data can be automatically drawn from a source URL?  
This would let people spend more time focusing on the fields that can't be inferred (what a source is used for, what other scripts / datasets it is used with)

