What is the Plan-and-Execute agent?
Think about how you cook a meal. First, you decide on a recipe, gather the ingredients, and then start cooking. Similarly, the Plan-and-Execute agent works by first planning its actions and then executing them.
How does it work?
The process is split into two main parts:
- The Planner: This is where the AI thinks about the task you've given it. It processes the information, makes a plan, and gives you a list of steps, much like how a route planner works.
- The Executor: After the planning is done, the executor takes over. It follows the plan and decides on the best tools or plugins to achieve the goal.
Benefits and considerations
- Focused approach: By separating planning and execution, the agent can concentrate better on each phase. This means it's more reliable.
- Flexibility: If we ever need to change or improve parts of the agent, it's easier because of this separated approach.
- Efficiency: While it might need more interactions to finish a task, it can use faster and more cost-effective models to get the job done.
How to use the Plan-and-Execute agent on Spell?
Using the agent is straightforward. When you input a task, the agent will first plan out the steps and then execute them.
To create an agent of the Plan-and-Execute type, make sure to choose it in the dropdown during the agent creation:
