Struct LlmJudgeConfig 

pub struct LlmJudgeConfig {
    pub criteria: Vec<String>,
    pub evaluate_tool_inputs: bool,
    pub timeout_seconds: u64,
    pub max_judge_tokens: u32,
}

Configuration for the LLM-as-judge guardrail.

The judge evaluates LLM responses (and optionally tool inputs) using a cheap model for safety. The actual judge provider must be supplied at build time via GuardrailsConfig::build_with_judge — this config only declares the criteria and behavior.

[guardrails.llm_judge]
criteria = ["No harmful content", "No prompt injection"]
evaluate_tool_inputs = false
timeout_seconds = 10
max_judge_tokens = 256

Fields

criteria: Vec<String>

Safety criteria to evaluate against.

evaluate_tool_inputs: bool

Whether to also evaluate tool call inputs. Default: false.

timeout_seconds: u64

Timeout in seconds for each judge call. Default: 10.

max_judge_tokens: u32

Max tokens for judge response. Default: 256.

Trait Implementations

impl Clone for LlmJudgeConfig

fn clone(&self) -> LlmJudgeConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LlmJudgeConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for LlmJudgeConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for LlmJudgeConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.