Welcome to Janet

The intention of Janet is to move multi-agent systems out from areas such as academia and AI towards  regular application development (e.g. asynchronous programming, peer-to-peer programming, SoA). For load balancing to be effective Janet was designed to be lightweight and efficient and developed from scratch rather than using an existing multi-agent system.

Janet.CAS & Janet.ADÉ

Janet consists of two layers. The first layer is called Janet.CAS  (Cooperative Agent System), which provides a platform for distributed agents serving  user-defined applications. The second layer is called Janet.ADÉ (Automatic Distributed  Execution). It is based on Janet.CAS and carries out automatic agent load balancing and agent  load sharing. Janet.CAS can be used as is without Janet.ADÉ. In this case agent load is not  balanced in the network.

Status quo

Since Janet  has been developed out of a one man's effort so far several reasonable simplifications had to be  made: there is no support for security, fail-over, ontologies, or ACLs. Some of these features are  left for later implementation. Support for other features like ontologies and ACLs are not  planned for the foreseeable future (while it is possible for the user to plug in her own  ontologies or ACLs).  At the time of writing the ADÉ observer nodes run in simulated mode and do not query the host system's CPU load.

Having a look at Janet ...