Custom backends

SteelThread is designed to allow for metrics to be pushed to other sinks, simply by implementing the correct metrics backend and passing it as config.

class StreamMetricsBackend(ABC):
    """Abstract interface for saving metrics."""

    @abstractmethod
    def save_metrics(self, metrics: list[StreamMetric]) -> None:
        """Save a list of tagged metrics for a specific evaluation run.

        Args:
            metrics (list[StreamMetricWithTags]): The metrics to save.

        """
        raise NotImplementedError


class MyMetricsBackend(StreamMetricsBackend):
    def save_metrics(self, metrics: list[StreamMetric]) -> None:
        return    

conf = StreamConfig(stream_name="stream_v1", config=config, metrics_backends=[MyMetricsBackend()])