Building action models with agents
Agents are the newest feature in the generative AI world. They are powerful tools that enable the automation of tasks by allowing generative AI models to take actions on our behalf. They act as intermediaries between the generative AI models and external systems or services, facilitating the execution of tasks in the real world.
Under the hood, an agent “understands” what the user wants and calls a backend function that performs the action. The scope within which the agent can act is defined by an OpenAPI schema that it will use both to “understand” what it does and how to properly call the backend function.
So, in summary, to build an agent we need an OpenAPI schema, a backend function, and a knowledge base. The knowledge base is optional, but it can greatly improve a user’s experience with the AI assistant.
For this section’s exercise, we will build an agent that “knows” the available...