crackerjack 0.37.9__py3-none-any.whl → 0.45.2__py3-none-any.whl

crackerjack/services/quality/anomaly_detector.py

@@ -0,0 +1,392 @@
+"""ML-based anomaly detection service for quality metrics analysis."""
+
+import logging
+import statistics
+import typing as t
+from collections import defaultdict, deque
+from dataclasses import dataclass, field
+from datetime import datetime
+from pathlib import Path
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class MetricPoint:
+    """Individual metric data point."""
+
+    timestamp: datetime
+    value: float
+    metric_type: str
+    metadata: dict[str, t.Any] = field(default_factory=dict[str, t.Any])
+
+
+@dataclass
+class AnomalyDetection:
+    """Anomaly detection result."""
+
+    timestamp: datetime
+    metric_type: str
+    value: float
+    expected_range: tuple[float, float]
+    severity: str  # low, medium, high, critical
+    confidence: float
+    description: str
+    metadata: dict[str, t.Any] = field(default_factory=dict[str, t.Any])
+
+
+@dataclass
+class BaselineModel:
+    """Statistical baseline model for a metric."""
+
+    metric_type: str
+    mean: float
+    std_dev: float
+    min_value: float
+    max_value: float
+    sample_count: int
+    last_updated: datetime
+    seasonal_patterns: dict[str, float] = field(default_factory=dict[str, t.Any])
+
+
+class AnomalyDetector:
+    """ML-based anomaly detection system for quality metrics."""
+
+    def __init__(
+        self,
+        baseline_window: int = 100,
+        sensitivity: float = 2.0,
+        min_samples: int = 10,
+    ):
+        """Initialize anomaly detector.
+
+        Args:
+            baseline_window: Number of recent samples for baseline calculation
+            sensitivity: Standard deviation multiplier for anomaly threshold
+            min_samples: Minimum samples required before anomaly detection
+        """
+        self.baseline_window = baseline_window
+        self.sensitivity = sensitivity
+        self.min_samples = min_samples
+
+        # Data storage
+        self.metric_history: dict[str, deque[MetricPoint]] = defaultdict(
+            lambda: deque[MetricPoint](maxlen=baseline_window)
+        )
+        self.baselines: dict[str, BaselineModel] = {}
+        self.anomalies: list[AnomalyDetection] = []
+
+        # Configuration
+        self.metric_configs = {
+            "test_pass_rate": {"critical_threshold": 0.8, "direction": "both"},
+            "coverage_percentage": {"critical_threshold": 0.7, "direction": "down"},
+            "complexity_score": {"critical_threshold": 15.0, "direction": "up"},
+            "execution_time": {"critical_threshold": 300.0, "direction": "up"},
+            "memory_usage": {"critical_threshold": 1024.0, "direction": "up"},
+            "error_count": {"critical_threshold": 5.0, "direction": "up"},
+        }
+
+    def add_metric(
+        self,
+        metric_type: str,
+        value: float,
+        timestamp: datetime | None = None,
+        metadata: dict[str, t.Any] | None = None,
+    ) -> None:
+        """Add new metric point and update baseline."""
+        if timestamp is None:
+            timestamp = datetime.now()
+
+        point = MetricPoint(
+            timestamp=timestamp,
+            value=value,
+            metric_type=metric_type,
+            metadata=metadata or {},
+        )
+
+        self.metric_history[metric_type].append(point)
+
+        # Update baseline if we have enough samples
+        if len(self.metric_history[metric_type]) >= self.min_samples:
+            self._update_baseline(metric_type)
+
+            # Check for anomalies
+            anomaly = self._detect_anomaly(point)
+            if anomaly:
+                self.anomalies.append(anomaly)
+                logger.info(f"Anomaly detected: {anomaly.description}")
+
+    def _update_baseline(self, metric_type: str) -> None:
+        """Update statistical baseline for a metric type."""
+        history = list[t.Any](self.metric_history[metric_type])
+        values = [point.value for point in history]
+
+        # Calculate basic statistics
+        mean = statistics.mean(values)
+        std_dev = statistics.stdev(values) if len(values) > 1 else 0
+        min_val = min(values)
+        max_val = max(values)
+
+        # Detect seasonal patterns (hourly, daily)
+        seasonal_patterns = self._detect_seasonal_patterns(history)
+
+        self.baselines[metric_type] = BaselineModel(
+            metric_type=metric_type,
+            mean=mean,
+            std_dev=std_dev,
+            min_value=min_val,
+            max_value=max_val,
+            sample_count=len(values),
+            last_updated=datetime.now(),
+            seasonal_patterns=seasonal_patterns,
+        )
+
+    def _detect_seasonal_patterns(self, history: list[MetricPoint]) -> dict[str, float]:
+        """Detect seasonal patterns in metric history."""
+        patterns: dict[str, float] = {}
+
+        if len(history) < 24:  # Need at least 24 points for pattern detection
+            return patterns
+
+        # Group by hour of day
+        hourly_values = defaultdict(list)
+        for point in history:
+            hour = point.timestamp.hour
+            hourly_values[hour].append(point.value)
+
+        # Calculate hourly averages
+        for hour, values in hourly_values.items():
+            if len(values) >= 3:  # Need at least 3 samples
+                patterns[f"hour_{hour}"] = statistics.mean(values)
+
+        return patterns
+
+    def _detect_anomaly(self, point: MetricPoint) -> AnomalyDetection | None:
+        """Detect if a metric point is anomalous."""
+        metric_type = point.metric_type
+        baseline = self.baselines.get(metric_type)
+
+        if not baseline:
+            return None
+
+        # Calculate expected range
+        lower_bound = baseline.mean - (self.sensitivity * baseline.std_dev)
+        upper_bound = baseline.mean + (self.sensitivity * baseline.std_dev)
+
+        # Apply seasonal adjustment if available
+        seasonal_adjustment = self._get_seasonal_adjustment(point, baseline)
+        if seasonal_adjustment:
+            lower_bound += seasonal_adjustment
+            upper_bound += seasonal_adjustment
+
+        # Check for anomaly
+        is_anomaly = point.value < lower_bound or point.value > upper_bound
+
+        if not is_anomaly:
+            return None
+
+        # Determine severity
+        severity = self._calculate_severity(point, baseline, lower_bound, upper_bound)
+
+        # Calculate confidence
+        confidence = self._calculate_confidence(point, baseline)
+
+        # Generate description
+        description = self._generate_anomaly_description(
+            point, baseline, lower_bound, upper_bound, severity
+        )
+
+        return AnomalyDetection(
+            timestamp=point.timestamp,
+            metric_type=metric_type,
+            value=point.value,
+            expected_range=(lower_bound, upper_bound),
+            severity=severity,
+            confidence=confidence,
+            description=description,
+            metadata=point.metadata,
+        )
+
+    def _get_seasonal_adjustment(
+        self, point: MetricPoint, baseline: BaselineModel
+    ) -> float:
+        """Get seasonal adjustment for the current time."""
+        hour = point.timestamp.hour
+        hour_pattern = baseline.seasonal_patterns.get(f"hour_{hour}")
+
+        if hour_pattern is not None:
+            return hour_pattern - baseline.mean
+
+        return 0.0
+
+    def _calculate_severity(
+        self,
+        point: MetricPoint,
+        baseline: BaselineModel,
+        lower_bound: float,
+        upper_bound: float,
+    ) -> str:
+        """Calculate anomaly severity based on deviation magnitude."""
+        if baseline.std_dev == 0:
+            return "medium"
+
+        # Check for critical threshold breaches first
+        if self._is_critical_threshold_breached(point):
+            return "critical"
+
+        # Calculate z-score and map to severity
+        z_score = self._calculate_z_score(point, baseline, lower_bound, upper_bound)
+        return self._severity_from_z_score(z_score)
+
+    def _is_critical_threshold_breached(self, point: MetricPoint) -> bool:
+        """Check if point breaches critical thresholds."""
+        config = self.metric_configs.get(point.metric_type, {})
+        critical_threshold = config.get("critical_threshold")
+
+        if not critical_threshold:
+            return False
+
+        direction = config.get("direction", "both")
+        threshold_float: float = (
+            float(str(critical_threshold)) if critical_threshold is not None else 0.0
+        )
+        return self._threshold_breached_in_direction(
+            point.value, threshold_float, str(direction)
+        )
+
+    def _threshold_breached_in_direction(
+        self, value: float, threshold: float, direction: str
+    ) -> bool:
+        """Check if value breaches threshold in specified direction."""
+        if direction == "up":
+            return value > threshold
+        elif direction == "down":
+            return value < threshold
+        elif direction == "both":
+            return value > threshold or value < -threshold
+        return False
+
+    def _calculate_z_score(
+        self,
+        point: MetricPoint,
+        baseline: BaselineModel,
+        lower_bound: float,
+        upper_bound: float,
+    ) -> float:
+        """Calculate z-score for the point."""
+        deviation = min(abs(point.value - lower_bound), abs(point.value - upper_bound))
+        return deviation / baseline.std_dev
+
+    def _severity_from_z_score(self, z_score: float) -> str:
+        """Map z-score to severity level."""
+        if z_score > 4:
+            return "critical"
+        elif z_score > 3:
+            return "high"
+        elif z_score > 2:
+            return "medium"
+        return "low"
+
+    def _calculate_confidence(
+        self, point: MetricPoint, baseline: BaselineModel
+    ) -> float:
+        """Calculate confidence in anomaly detection."""
+        # Base confidence on sample size and consistency
+        sample_factor = min(baseline.sample_count / 50, 1.0)  # Max at 50 samples
+
+        # Factor in standard deviation consistency
+        if baseline.std_dev == 0:
+            std_factor = 0.5  # Low confidence for constant values
+        else:
+            # Higher confidence for more consistent baselines
+            cv = baseline.std_dev / abs(baseline.mean) if baseline.mean != 0 else 1
+            std_factor = max(0.1, min(1.0, 1.0 - cv))
+
+        return sample_factor * std_factor
+
+    def _generate_anomaly_description(
+        self,
+        point: MetricPoint,
+        baseline: BaselineModel,
+        lower_bound: float,
+        upper_bound: float,
+        severity: str,
+    ) -> str:
+        """Generate human-readable anomaly description."""
+        direction = "above" if point.value > upper_bound else "below"
+        expected_range = f"{lower_bound:.2f}-{upper_bound:.2f}"
+
+        return (
+            f"{severity.title()} anomaly in {point.metric_type}: "
+            f"value {point.value:.2f} is {direction} expected range "
+            f"{expected_range} (baseline: {baseline.mean:.2f})"
+        )
+
+    def get_anomalies(
+        self,
+        metric_type: str | None = None,
+        severity: str | None = None,
+        since: datetime | None = None,
+        limit: int = 100,
+    ) -> list[AnomalyDetection]:
+        """Get filtered anomalies."""
+        anomalies = self.anomalies
+
+        # Apply filters
+        if metric_type:
+            anomalies = [a for a in anomalies if a.metric_type == metric_type]
+
+        if severity:
+            anomalies = [a for a in anomalies if a.severity == severity]
+
+        if since:
+            anomalies = [a for a in anomalies if a.timestamp >= since]
+
+        # Sort by timestamp (newest first) and limit
+        anomalies.sort(key=lambda x: x.timestamp, reverse=True)
+        return anomalies[:limit]
+
+    def get_baseline_summary(self) -> dict[str, dict[str, t.Any]]:
+        """Get summary of all baseline models."""
+        summary = {}
+
+        for metric_type, baseline in self.baselines.items():
+            summary[metric_type] = {
+                "mean": baseline.mean,
+                "std_dev": baseline.std_dev,
+                "range": (baseline.min_value, baseline.max_value),
+                "sample_count": baseline.sample_count,
+                "last_updated": baseline.last_updated.isoformat(),
+                "seasonal_patterns": len(baseline.seasonal_patterns),
+            }
+
+        return summary
+
+    def export_model(self, output_path: str | Path) -> None:
+        """Export anomaly detection model for persistence."""
+        import json
+
+        model_data = {
+            "baselines": {
+                metric_type: {
+                    "metric_type": baseline.metric_type,
+                    "mean": baseline.mean,
+                    "std_dev": baseline.std_dev,
+                    "min_value": baseline.min_value,
+                    "max_value": baseline.max_value,
+                    "sample_count": baseline.sample_count,
+                    "last_updated": baseline.last_updated.isoformat(),
+                    "seasonal_patterns": baseline.seasonal_patterns,
+                }
+                for metric_type, baseline in self.baselines.items()
+            },
+            "config": {
+                "baseline_window": self.baseline_window,
+                "sensitivity": self.sensitivity,
+                "min_samples": self.min_samples,
+            },
+            "exported_at": datetime.now().isoformat(),
+        }
+
+        with open(output_path, "w", encoding="utf-8") as f:
+            json.dump(model_data, f, indent=2)