Skip to main content

Core API

CortexScheduler is a key class within the SocialAGI library. It operates as a task or MentalProcess manager for an AI, overseeing a collection of mental processes or "jobs", and determining their order of execution.

Adhering to the principles of concurrency control, CortexScheduler ensures that no two jobs interfere with each other, providing a seamless and safe task management mechanism.

Key Concepts and Methods​

  1. Initialization of CortexScheduler

Initialize CortexScheduler by providing an initial CortexStep. If you want to customize job ordering, you can optionally supply a QueuingStrategy.

const cortexScheduler = new CortexScheduler(firstStep);
// Or with a custom QueuingStrategy
const cortexScheduler = new CortexScheduler(firstStep, {
  queuingStrategy: customQueuingStrategy,
});
  1. Registration of Mental Processes with register()

Use the register() method to add a new mental process. It requires a ProcessConfig object that includes the process name and the process function.

cortexScheduler.register({
  name: "ProcessName",
  process: async (signal, newMemory, lastStep) => {
    // Implement your process here
  },
});
  1. Execution of a Mental Process via dispatch()

The dispatch() method schedules a registered mental process to be executed when resources are next available. It takes the process name and a newMemoryobject as parameters. The newMemory object is a ChatMessage type, representing the current user message or a system-generated message. The method returns aPromise that resolves when the job is complete.

// newMemory is of type ChatMessage
const newMemory = { role: "user", content: "Hello, AI!" };
cortexScheduler
  .dispatch("ProcessName", newMemory)
  .then(() => {
    // Actions to perform after the job is done
  })
  .catch((error) => {
    // Error handling
  });

Queuing Strategy​

CortexScheduler uses a default queuing strategy that adds new jobs to the end of the job queue. To employ a custom strategy, supply a QueuingStrategy function during initialization. This function dictates the placement of a new job in the queue, taking into account the current job and the existing queue.

const customQueuingStrategy = (currentJob, queue, newJob) => {
  // Define your custom queuing strategy
};

const cortexScheduler = new CortexScheduler(firstStep, {
  queuingStrategy: customQueuingStrategy,
});

Managing Jobs​

In the CortexScheduler context, each mental process is a job. The scheduler manages these jobs according to the queuing strategy, ensuring their orderly execution. This offers an efficient way to control multiple mental processes in AI applications.