When creating or editing an Agent, three tabs are shown. The first tab, titled Configuration, is divided into two main panels: Agent Details and Prompt.
This section allows you to define the core identity of your agent. It includes the following fields:
- Agent Name (required): This field is used to assign a clear and recognizable name to your Agent. This name is used to identify the Agent throughout the platform.
- Agent Purpose (required): Here you define the main purpose of the Agent; that is, what kind of task it solves or what assistance it provides to the end user. This description guides the Agent's behavior and provides it with the necessary context to respond accurately. For example, if the agent is intended to provide flight cancellation support, that should be its purpose.
- Agent Role: This field complements the Purpose by describing the Agent's role in a short phrase, almost like a job title. While the Purpose indicates what the Agent does, the Role reflects who the Agent is or how the Agent presents itself. For example, if the purpose is to help with cancellations, the role might be "Airline Customer Service Assistant".
- Avatar image: Based on the title and description provided, Globant Enterprise AI Lab can automatically generate an avatar. By clicking on Generate image, you will see a modal with different avatar options. You can choose one or click on Regenerate to generate a new set. Once you're satisfied, click on Accept to assign the selected avatar to your agent.
This section establishes the key elements that define how the Agent should behave during its interactions. It is organized in three main fields:
- Background Knowledge: Here you can include contextual information or essential knowledge that the Agent must be aware of in order to answer correctly. This can include product descriptions, internal company processes or any other relevant data. This information is used by the model to generate more accurate answers aligned with the Agent's specific domain.
- Guidelines: In this field you define clear instructions on how the Agent should behave in different scenarios. For example, if there is an expected tone of voice, mandatory steps before giving a response or policies to follow. These guidelines help to maintain consistency in the responses and ensure that the Agent acts in a way that is aligned with the project objectives.
Note: It is mandatory to complete at least one of the two fields (Background Knowledge or Guidelines) since they are essential to define what the Agent should do and how it should respond.
- Examples: This section allows you to enter input and output data pairs to illustrate how the agent is expected to respond to certain messages. To add a new example, click on + Add new example. This will open a row with two columns: one for the input message (Input Data) and one for the expected response from the agent (Output Data).
To delete an example, click on the Delete icon. This will activate the Confirm and Cancel icons, which allow you to confirm or cancel the deletion, respectively.
Since April 2025 release.