Agents

Agents in Braintrust allow you to chain together two or more prompts. You can create or edit agents in the playground, and view and execute them from the library.

Creating an agent in the playground

To create an agent, navigate to a playground and select +Agent. Start by creating the base prompt or selecting one from your library. Then, create or select another prompt by selecting the + icon in the comparison agent pane.

The prompts will chain together, and they will automatically run consecutively.

Variables

When building agents, the first prompt node and subsequent prompt nodes have slightly different templating behavior. To refer to variables, you can use mustache templating syntax.

Using dataset

The dataset variable is globally available. You can use it in any agent prompt node to access input, expected, and metadata. For example, to access metadata.foo, use {{dataset.metadata.foo}}.

In the first agent prompt

The first prompt node can access dataset variables directly by using {{input}}, {{expected}}, and {{metadata}}, for convenience. Using {{dataset}} is useful for consistent access patterns with later prompts, or if you use this prompt in a later position in the future, since all agent prompts can access {{dataset}}.

Later prompts

Subsequent prompts can access the output of the previous node by using {{input}}.

• If the previous node outputs structured data, use dot notation. For example, {{input.bar}}.
• If the previous node outputs text or unschematized JSON, you can only use {{input}}. If you're using JSON outputs, consider switching to structured outputs to enable accessing nested output variables with linting and autocomplete.

Viewing and executing agents

You can view and execute single runs of agents from your agent library, but you will not be able to edit them or see them run.