Designing a REST API with redundant data: client-side or server-side data processing - rest

Question

I'm developing a JSON-based REST API allowing to get information about documents where a 'document' resource has the following form:
{
"id": 1,
"name": "document 1",
...
"fields":
[
{"name": "field1", "category": "category1", ...},
{"name": "field2", "category": "category2", ...},
...
]
}
The GET /documents/:id route is quite straight-forward, and I want the GET /documents route to provide an array of basic information related to documents instead of just links or IDs only, something like:
[
{"id": 1, "name": "document 1"},
...
]
Now the UI needs to display the list of documents with the list of categories the fields of a document belong to:
document 1 (category1, category2)
document 2 (category2, category3, category4)
...
The first solution would be to add the fields field to each document in the response of GET /documents and let the client compute the list of categories (maybe this could lead to a poor UX when there are a lot of documents to display), eg:
[
{"id": 1, "name": "document1", "fields":
[
{"name": "field1", "category": "category1", ...},
...
]
]
The second one would be to add that piece of (redundant) information server-side to avoid to add too much data in the response and avoid the client to loop through each document's fields (but now the API is more dependent on how the UI presents the data), eg:
[
{"id", "name": "document 1", "categories": ["category1", "category2"]},
...
]
According to your experiences, which solution should I use?