As of 48af5dd, the API for creating agents that operate in a sequence is:

Dispatcher().chain(initial_event, [sequence of agents])

For example, in the chunker demo:

dispatcher.chain("button" , Chunker(5), SimLampHandler("http://localhost:9090/"))

This is a vast improvement on the previous API using just the constructors of the agents, but is still a bit simplistic.

One significant issue is that currently, if you make two seperate chains with a common start, there is no actual sharing of agents. This could be addressed by adding a sort of "value hash" to the agents, where two separate agent objects with the same parameters would hash to the same value, preventing them from being duplicated in the agent manager.

Another issue is that the current API doesn't easily allow for expressing branches in the event flow, even if the above solution is implemented to avoid duplicate agents. How should we best solve this?