Interface: TextGuardrailReport
Defined in: packages/agentos/src/api/runtime/textGuardrails.ts:81
What each stage's guards answered on one call, as codes and ids.
Properties
input
input:
ModerationEvaluation[]
Defined in: packages/agentos/src/api/runtime/textGuardrails.ts:83
mode
mode:
"enforce"|"dryRun"
Defined in: packages/agentos/src/api/runtime/textGuardrails.ts:82
output
output:
ModerationEvaluation[]
Defined in: packages/agentos/src/api/runtime/textGuardrails.ts:86
result
result:
"stopped"|"passed"|"sanitized"|"replaced"
Defined in: packages/agentos/src/api/runtime/textGuardrails.ts:91
passed: nothing changed. sanitized: a guard rewrote the input or the reply. replaced: a fixed reply took its
place. stopped: the call ended in a GuardrailStopError, whose report this is.
toolCalls
toolCalls:
object[]
Defined in: packages/agentos/src/api/runtime/textGuardrails.ts:85
toolCallId and toolName as on TextGuardrailVerdict: the module's number and a registered name or unregistered.
evaluations
evaluations:
ModerationEvaluation[]
toolCallId
toolCallId:
string
toolName
toolName:
string