Interface: ModerationDecision
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:44
Properties
deciding?
optionaldeciding:ModerationEvaluation
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:56
The evaluation that decided: a rule's BLOCK, else a hold, else a closed guard's failure, else a FLAG.
enforced
enforced:
boolean
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:49
evaluations
evaluations:
ModerationEvaluation[]
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:58
Every evaluation, in registration order.
hold?
optionalhold:GuardrailHold
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:60
Set when the outcome, or wouldBe, is hold.
outcome
outcome:
ModerationOutcome
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:46
What the host does. In a dry run this is always allow.
sanitizedText?
optionalsanitizedText:string
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:66
In enforce: the text as the sanitizers left it, when a guard with canSanitize changed it, whatever the outcome;
'' when one removed it (modifiedText: null). It is the person's text, so a host keeps it out of its logs. Never
set in a dry run.
unavailable
unavailable:
boolean
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:54
Nothing but a guard that threw, timed out or answered nonsense, or an image the image guard could not have checked, kept the content back: no rule refused it and no guard held it.
wouldBe?
optionalwouldBe:ModerationOutcome
Defined in: packages/agentos/src/safety/guardrails/contentModeration.ts:48
In a dry run: the outcome enforcement would have given.