So far, we have examined how LLM agents are defined and how they utilize core components like Large Language Models, tools, and memory. We've also walked through creating a very basic agent. Now, we turn our attention to a more dynamic aspect of agent behavior: how they decide what to do and in what order. For an agent to be truly useful beyond single-step operations, it needs a mechanism to formulate and follow a plan to achieve a given objective.

This chapter introduces foundational approaches to agent planning and execution. You will learn about:

• The general concept of planning in the context of LLM agents and its importance for completing tasks that require multiple steps.
• Techniques like Chain-of-Thought prompting, which helps guide the agent's reasoning process.
• The ReAct (Reason and Act) framework, a common pattern that combines an agent's reasoning capabilities with its ability to take actions.
• How to set clear objectives for your agent and decompose more involved tasks into simpler, sequential actions.
• Basic methods for observing an agent's progress and managing common, simple failures during execution.

By working through these topics, you'll gain an understanding of how to structure an agent's thinking process, enabling it to approach and carry out sequences of actions to reach a specified goal.