Enum ResponseDecision 

#[non_exhaustive]
pub enum ResponseDecision {
    Accept,
    Block(String),
    RetryWithFeedback(String),
}

Decision returned by AgentHooks::on_llm_response — an output guardrail that runs after the provider responds but before the response is persisted and surfaced.

This is the place for output moderation or secret-leakage detection.

Forward-compat is fail-open: the agent loop matches this enum with a wildcard arm that treats unknown variants as Accept, so a restrictive variant added here in a future release is silently ignored by loops built against older SDKs. Never add a variant whose ignored meaning weakens a security decision — extend Block/RetryWithFeedback semantics instead, or plan a breaking change.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Accept

Accept the response as-is.

Block(String)

Reject the response; the string explains why.

RetryWithFeedback(String)

Reject the response and feed the string back to the model so it can retry on the next turn.

Trait Implementations

impl Clone for ResponseDecision

fn clone(&self) -> ResponseDecision

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ResponseDecision

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.