When using dgeni we currently process two distinct types of documents: js code with jsdoc style tags and bower.json files (we use those to create a Dependencies index page where you can see links to the documentation of all the projects we use).

Currently we kind of hacked that capability in (by simulating the output of jsdoc for the JSON files). But really we would want different kind of pipelines that would converge at some point. So we would want something like this:

read js files ---- process jsdoc tags -
                                        \
                                         >--- render templates --- write files
                                        /
read json files --- process info -------