Enum LlmClassifierConfig 

#[non_exhaustive]
pub enum LlmClassifierConfig {
    Capability {
        judge_target: LlmTarget,
        efficient_target: LlmTarget,
        capable_target: LlmTarget,
        config: TaskClassifierConfig,
    },
    Escalation {
        judge_target: LlmTarget,
        efficient_target: LlmTarget,
        capable_target: LlmTarget,
        contract: ClassifierContractConfig,
        config: EscalationJudgeConfig,
        max_output_tokens: u64,
    },
    Custom {
        judge_target: LlmTarget,
        targets: Vec<(String, LlmTarget)>,
        default_target: String,
        config: CustomClassifierConfig,
    },
}

Complete construction settings for one LLM classifier mode.

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.

Capability

Routes between efficient and capable targets from a task-level verdict.

Fields

judge_target: LlmTarget

Target that produces classifier verdicts.

efficient_target: LlmTarget

Target used when the efficient tier can handle the task.

capable_target: LlmTarget

Target used when the task needs the capable tier.

config: TaskClassifierConfig

Capability classifier settings.

Escalation

Judges efficient responses and escalates after a confirmed streak.

Fields

judge_target: LlmTarget

Target that produces escalation verdicts.

efficient_target: LlmTarget

Target called before each escalation decision.

capable_target: LlmTarget

Target used after escalation is confirmed.

contract: ClassifierContractConfig

Prompt and verdict contract settings for the escalation judge.

config: EscalationJudgeConfig

Escalation policy settings.

max_output_tokens: u64

Maximum completion tokens available to the escalation verdict.

Custom

Routes among named targets using a user-supplied schema and policy.

Fields

judge_target: LlmTarget

Target that produces classifier verdicts.

targets: Vec<(String, LlmTarget)>

User-facing labels paired with their resolved routing targets.

default_target: String

Label selected when the judge does not produce a usable verdict.

config: CustomClassifierConfig

Custom classifier settings.