Logging Enhancements

Status: Complete

Project-wide structured logging configuration for local development and cloud deployments.

Quick Start

from src.logger import get_logger, configure_logging

configure_logging()
logger = get_logger(__name__)
logger.info("Application started")

Configuration

The logger auto-detects the deployment environment:

configure_logging()
configure_logging(level="INFO", use_json=True)
configure_logging(level="DEBUG", use_json=False)

Environment Auto-Detection

The logger checks for cloud hosting indicators:

Environment	Detection Variable
Azure App Service	WEBSITE_INSTANCE_ID
Azure Functions	AZURE_FUNCTIONS_ENVIRONMENT
AWS Lambda	AWS_EXECUTION_ENV
Kubernetes	KUBERNETES_SERVICE_HOST

Structured Logging

Add queryable custom fields:

logger.info(
    "Model training complete",
    extra={'extra_fields': {
        'model_name': 'XGBClassifier',
        'f1_score': 0.843
    }}
)

Logging Levels

• DEBUG: Detailed diagnostic information
• INFO: General operational events
• WARNING: Something unexpected but not critical
• ERROR: Serious problem; application continues
• CRITICAL: Application failure

Modules Enhanced

Module	Purpose
titanic_ml/data/loader.py	Data loading logging
titanic_ml/data/transformer.py	Transformation logging
titanic_ml/features/build_features.py	Feature engineering logging
titanic_ml/models/train.py	Training metrics with structured data
titanic_ml/models/predict.py	Prediction tracking with metrics
titanic_ml/app/routes.py	API request/response logging

Best Practices

1. Use structured logging for metrics (avoid free-form text)
2. Log at appropriate levels (INFO for operations, DEBUG for diagnostics)
3. Include context (model names, feature counts, data shapes)
4. Avoid logging sensitive data (PII, credentials, passwords)
5. Log summaries not entire objects

---

For cloud deployments, configure environment-specific logging levels and ensure log aggregation is enabled.