DriftGuard 🛡️

Enterprise-grade real-time ML model drift monitoring.

DriftGuard is a professional Python library designed for MLOps Engineers and Data Scientists to detect Data Drift and Concept Drift in production environments.

Features

• Statistical Drift Detection: Built-in support for Kolmogorov-Smirnov Test and Population Stability Index (PSI).
• Extensible: Easy to create custom detectors by inheriting from BaseDetector.
• Alerting System: Built-in Webhook integration to send real-time alerts to Slack, Microsoft Teams, etc.

Installation

You can install DriftGuard locally:

pip install -e .

Quick Start

import numpy as np
from driftguard.detectors.ks_drift import KSDriftDetector

# 1. Reference Data (e.g. from your Training Set)
reference_data = np.random.normal(0, 1, 1000)

# 2. Initialize and fit the detector
detector = KSDriftDetector(threshold=0.05)
detector.fit(reference_data)

# 3. New Data (e.g. Production Data)
current_data = np.random.normal(1.5, 1.5, 1000)

# 4. Detect Drift
report = detector.detect(current_data)

print(report)
# {0: {'detector': 'Kolmogorov-Smirnov Test', 'drift_detected': True, 'metrics': {'ks_statistic': ..., 'p_value': ...}, 'threshold': 0.05}}

Running Tests

To run the unit tests:

pip install pytest
pytest tests/