Currently, both formats are supported for backwards compatibility

tested briefly in types

- If Query is a CONSTRUCT and client requests RDF/XML, TTL, NT, JSON -> works as usual
- if query is a CONSTRUCT and client requests HTML
-- Results hang from 'rdf', similar to 'models'

```
{%for s in rdf.subjects%}
  <li>{{s.value}}</li>
{%endfor%}
```

Returns all the subjects. `{{s.value}}`, `{{s.curie}}`, etc.  work the same way as in models

```
{%for p in rdf.ex__alvaro%}
  {{p.value}}
{%endfor%}

```

Returns all the predicates from `ex:alvaro` (the colon is translated to `__`). Assuming the namespace for `ex` is `http://example.org/`, this means all the rpedicates from `http://example.org/alvaro`.

```
{%for o in rdf.ex__alvaro.rdfs__label%}
  {{o.value}}
{%endfor%}

```

returns all the objects from triples <http://example.org/alvaro> rdfs:label ?o

TODO:

- Blank nodes is unclear that are working now
- Language, datatypes, etc. non-existant
- Discussion: Combine all the models into one graph? or name them similar to 'models'? Currently adding more queries crashes
- Probably a bunch of bugs

GUI implemented

TODO: GUI needs work

This decision was made based on the lack of CORS-enabled SPARQL endpoints, which difficults the use of client-base queries.

Adding explanation when sqlite db is not writable