...
The following use cases are what prompted the development of the tool (note, while all the uses cases given below are SAML-focused the general aggregator product is not SAML-focused).
A command line tool that can read in a bunch of SAML entity descriptors, schema validate them, check certain policy constraints, filter out entities based on given rules, assemble the entities into multiple different SAML metadata documents, and sign them all.
A command line tool and web service that can perform the technical work of inter-federation. This essentially means doing everything in the previous use case but operating on whole metadata documents from various federations.
A web service that can operate as a metadata "oracle" for IdPs and SPs and hide the complexity of fetching, validating, and munging (potentially many) metadata sources. This service could be deployed within an organization and the organization's IdPs and SPs could delegate most of the hard work of dealing with metadata to the oracle, making their configuration simpler, decreasing network load, and potentially increasing resiliency in cases where external metadata may go offline.
The aggregator itself is a fairly general tool and so it's quite likely that as adoption grows people will find other uses for it as well. It is certainly the goal of the developers to make it relatively easy to meet any use case with a general form of "read in a bunch of data, transform it in various ways, and write it out or search it".
...
The web service is built of:
a generative pipeline that is responsible for reading in all the data and creating the item collection that can be searched
an item store that is responsible for holding on to, indexing, and querying the item collection created by the generative pipeline
a result processing pipeline which is responsible for performing per-query work to prepare the query results to be sent back to the requester (e.g., merging multiple items in to a single document, signing the result, etc.)
a query controller which receives query requests, gets queries the item store, runs the results through the result processing pipeline, and sends the response back (i.e., this components coordinates the others)
...