Set up a server for hosting AI generated IIIF manifests

[jacobthill]

We want to use AI to generate a table of manuscript illuminations, table of contents, etc. We don't have the ability to overwrite existing IIIF manifests. The workflow according to this slack discussion would be to generate a new "AI annotated" manifest and link to the canonical manifest in the AI manifest. We would then need to host the AI manifest. In the AI manifest we should add a link to the canonical manifest and also a structures section like in this recipe.

The manifest url needs to follow a predicable pattern that uses the dlme-record-id e.g. www.example.com/{#dlme-record-id}/manifest. This way we can put the manifest url in the intermediate representation before the manifest exists.

We will also need the canonical url from the data provider to build the new manifest so we may need to add a canonical_manifest field to the IR. If we do that we could:

• Put the link to the new AI manifest in the IR during transform before the new AI manifest exists.
• Put the canonical_manifest url in the IR and use that post transform to generate the new AI manifest.
• Generate AI metadata to populate the IR and the new AI manifest post transform.