For more information, I recommend reading up on Wikipedia: Microservices
Intelligence, reasoning, and learning are all huge, complex tasks. It’s too big and complex for a single human, or even a single team of humans, to comprehend. You can spend a lifetime studying just one domain, such as visual processing, and still not master it. Because of this complexity, AGI will be achieved by collaboration between many domain experts. Each aspect of intelligence can be worked on by individual teams and experts. MARAGI allows for the creation of a pluggable architecture, where new microservices can be easily added to the greater whole.
MARAGI microservices all have two basic functions: input and output. This simplicity is reflected in the design of the MARAGI client and server. The third, hidden function is processing. Each microservice should process the information it takes in and produce something novel as part of the output.
The input of a microservice can be an external device, such as a camera or microphone. Microservices can also consume the output of other microservices. Some microservices are dedicated to interacting with sensors, such as cameras, microphones, orientation, acceleration, and any other kind of hardware sensor. Other microservices rely on the input from other microservices. For instance, an object detection microservice relies on the input from camera microservices.
The output of microservices can be deep learning inferences, raw data, evaluations, motor controls, or other output. Inference microservices consume data of some form and evaluate it with machine learning models. Hardware microservices take information from devices and translate it into JSON-friendly messages, such as raw video images or audio segments.
MARAGI, at its core, is meant to be an ecosystem of microservices. The success of MARAGI relies on the participation and contribution of independent programmers, researchers, students, and professionals.