Write your own LLM Eval class

💡

Have you seen our custom evaluation wrappers? They are a great way to quickly create your own evaluation class.

If you want to write your own evaluation class, you can do so by extending the BaseEvaluator class.

You can extend the BaseEvaluator class. See this example (opens in a new tab)

You need to implement the following methods:

@property
def _model(self):
    # Which model was used for this evaluation
    pass
 
@property
def name(self):
    return "UpperCamelCaseName"
 
@property
def display_name(self):
    return "Evaluation Display Name"
 
@property
def metric_ids(self) -> List[str]:
    return [MetricType.PASSED.value]
 
@property
def default_function_arguments(self):
    return {}
 
@property
def required_args(self):
    # expects an array of strings from ["query", "context", "response", "expected_response", "text"]
    pass
 
@property
def examples(self):
    pass
 
def is_failure(self) -> Optional[bool]:
    pass
 
 
def _evaluate(self, **kwargs) -> EvalResult:
    pass

💡

Contribute!

If you write an evaluator that could be useful for others, please consider raising a PR at https://github.com/athina-ai/athina-evals (opens in a new tab).

Cookbooks Develop Dashboard