Evaluator

gemini_adk_rs::evaluation

Trait Evaluator 

Source 
pub trait Evaluator: Send + Sync {
    // Required method
    fn evaluate<'life0, 'life1, 'life2, 'async_trait>(
        &'life0 self,
        actual: &'life1 [Invocation],
        expected: Option<&'life2 [Invocation]>,
    ) -> Pin<Box<dyn Future<Output = Result<EvalResult, EvalError>> + Send + 'async_trait>>
       where Self: 'async_trait,
             'life0: 'async_trait,
             'life1: 'async_trait,
             'life2: 'async_trait;
}
Expand description

Trait for evaluating agent invocations against expected results.

Mirrors ADK-Python’s Evaluator abstract class.

Required Methods§

Source

fn evaluate<'life0, 'life1, 'life2, 'async_trait>( &'life0 self, actual: &'life1 [Invocation], expected: Option<&'life2 [Invocation]>, ) -> Pin<Box<dyn Future<Output = Result<EvalResult, EvalError>> + Send + 'async_trait>>
where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait,

Evaluate agent invocations.

§Arguments
  • actual — The agent’s actual invocations.
  • expected — Optional expected (golden) invocations for comparison.
§Returns

An EvalResult containing per-metric and overall scores.

Implementors§