star-ray is an experimental multi-agent simulation platform that supports the development of AI agents and their environments.
Star stands for: Simulation Testbed for Agent Research, this platform has a long history of revisions (see e.g. pystarworlds), and was originally based on the GOLEM framework, it is conceptually similar to the previous instantiations but makes many practical improvements over previous iterations.
Ray is a unified framework for scaling AI and Python applications. Ray consists of a core distributed runtime and a set of AI libraries for simplifying [Machine Learning] compute.
Developing robust distributed systems is hard, rather than reinvent the wheel we decided to make use of Ray - a powerful distributed systems package that is widely used in the AI/ML community. It has a convenient API which is abstract enough to fit well with existing GOLEM concepts.
Star-ray provides abstractions that supports developers to quickly develop software (AI) agents and multi-agent simulations allowing you to focus on some of the more interesting aspects of AI agents such a their individual behaviour or group interaction.
Star-ray implements: sensors, actuators, an event system (actions and observations), and environment which can be extended to ease development. These provide the backbone to your simulation and abstract away from the complexities of parallel execution and distribution.
Official documentation is coming soon, if you wish to get started with star-ray before this check out the following package(s) for inspiration:
or contact the owners of this repo.