The XAgent system described here emphasizes autonomous decision-making in AI, promoting a structure that separates high-level planning from low-level execution, akin to human cognitive processes. This separation is facilitated through a dual-loop mechanism: the outer-loop for planning and the inner-loop for execution. Below, we delve into the logic of these loops and provide pseudo-code to illustrate their functions.
Outer-Loop: High-Level Planning
The outer-loop is responsible for strategizing and task segmentation. It initially creates a plan by dividing a complex task into simpler, actionable sub-tasks, then continuously refines this plan based on feedback from the execution of these sub-tasks.
Inner-Loop: Low-Level Execution
The inner-loop takes charge of the detailed execution of each sub-task. It identifies the necessary tools for the task, executes them, and then provides feedback to the outer-loop for potential plan refinements.
