Crate nous_judge 

Async LLM-as-judge evaluators for Nous.

These evaluators run asynchronously after agent runs complete. They use a separate model call to assess quality dimensions that require language understanding.

Re-exports

pub use anthropic_judge::AnthropicJudgeProvider;

pub use judge_provider::JudgeProvider;

pub use judge_provider::MockJudgeProvider;

pub use judge_provider::parse_judge_scores;

pub use plan_adherence::PlanAdherence;

pub use plan_quality::PlanQuality;

pub use task_completion::TaskCompletion;

Modules

anthropic_judge

Anthropic API-backed judge provider for real LLM-as-judge evaluation.

judge_provider

LLM call wrapper for evaluation.

plan_adherence

Plan adherence evaluator — did the agent follow its stated plan?

plan_quality

Plan quality evaluator — LLM-as-judge for reasoning coherence.

task_completion

Task completion evaluator — did the agent achieve its goal?