Integrations API Reference

MLflow Integration

assert_mlflow_model

Assert MLflow model properties and metrics.

from ml_assert.integrations.mlflow import assert_mlflow_model

# Assert MLflow model
assert_mlflow_model(
    model_uri="runs:/run_id/model",
    metrics={
        "accuracy": 0.8,
        "precision": 0.8,
        "recall": 0.8
    }
)

Parameters

• model_uri: MLflow model URI
• metrics: Dictionary of metric names and minimum values
• params: Dictionary of parameter names and expected values (optional)

Prometheus Integration

assert_prometheus_metrics

Assert Prometheus metrics for model monitoring.

from ml_assert.integrations.prometheus import assert_prometheus_metrics

# Assert Prometheus metrics
assert_prometheus_metrics(
    metrics={
        "model_accuracy": 0.8,
        "prediction_latency": 100
    },
    labels={"model": "my_model"}
)

Parameters

• metrics: Dictionary of metric names and thresholds
• labels: Dictionary of label names and values
• timeout: Maximum time to wait for metrics (default: 30 seconds)

Slack Integration

assert_slack_notification

Assert Slack notification delivery.

from ml_assert.integrations.slack import assert_slack_notification

# Assert Slack notification
assert_slack_notification(
    channel="#ml-monitoring",
    message="Model drift detected",
    timeout=30
)

Parameters

• channel: Slack channel name
• message: Expected message content
• timeout: Maximum time to wait for notification (default: 30 seconds)

DVC Integration

assert_dvc_artifact

Assert DVC artifact properties.

from ml_assert.integrations.dvc import assert_dvc_artifact

# Assert DVC artifact
assert_dvc_artifact(
    path="models/model.pkl",
    exists=True,
    size_mb=10
)

Parameters

• path: Path to DVC artifact
• exists: Whether artifact should exist
• size_mb: Expected size in megabytes (optional)
• md5: Expected MD5 hash (optional)