After you’ve created an agent, the model will show up alongside all the pre-built models in your project, and you can add it like any other model.

The agent will classify your content based on the safety categories you’ve defined when creating the agent.

Response signature

AI agents have a similar response signature to the pre-built classifiers, with the addition of a rule field.

label
string
required

The label of the rule that was triggered. Returns NEUTRAL if no rules were triggered.

score
number
required

The score of the most probable label. LLama Guard returns a probability score between 0 and 1, whereas GPT-4o-mini returns a binary classification 1 or 0.

label_scores
object
required

An object containing all the rules with either 0 or 1 as the score.

rule
string

The description of the rule that was triggered. Returns null if no rules were triggered.

Was this page helpful?