The folder contains support for flume to be used with Apex. It comprises mainly of two components. First is an agent that sits on the flume side, receives data from flume and makes it available via a socket server. In effect it converts a push to a pull model. The second component is the input operator that reads from the agent.
The project is started with the latest code at the time of the sub-module creation. For older history look at the flume sub-module in the older project called Megh (git@github.com:DataTorrent/Megh).