aeryl-sdk 0.1.2

Aeryl SDK for chaos testing and error detection

Aeryl SDK

A Python SDK for training and deploying machine learning models to detect and analyze chaos in production systems.

Installation

pip install aeryl-sdk

Quick Start

from aeryl_sdk import AerylModel

# Initialize and train a new model
model = AerylModel(
    dev_path='data/company_research_dev.csv',
    prod_path='data/company_research_prod.csv',
    models_dir='models'
)
model_id = model.train()

# Or load an existing model by ID
model = AerylModel(
    dev_path='data/company_research_dev.csv',
    prod_path='data/company_research_prod.csv',
    models_dir='models',
    model_id='your-model-id'
)

# Make predictions
predictions = model.predict()

AerylModel Class

The AerylModel class provides a high-level interface for training and inference using the chaos classifier.

Initialization

model = AerylModel(
    dev_path='path/to/dev/data.csv',    # Path to development dataset
    prod_path='path/to/prod/data.csv',  # Path to production dataset
    models_dir='models',                # Directory to save/load models
    model_id=None                       # Optional: ID of existing model to load
)

Key Methods

Training

# Train the model on the development dataset
model_id = model.train()

Returns a dictionary containing the model ID.

Loading a Model

# Load an existing model by ID
model.load_model(model_id)

This method is automatically called during initialization if a model_id is provided.

Prediction

# Make predictions on the production dataset
predictions = model.predict(model_id=None)  # Optional: specify model_id

Returns a dictionary containing predictions for each run and step:

{
    'model_id': 'uuid-123',
    'predictions': {
        'run_1': {
            'step_1': {'prediction': True, 'probability': 0.85},
            'step_2': {'prediction': False, 'probability': 0.12},
            'step_3': {'prediction': True, 'probability': 0.92}
        },
        'run_2': {
            'step_1': {'prediction': False, 'probability': 0.08},
            'step_2': {'prediction': False, 'probability': 0.15},
            'step_3': {'prediction': False, 'probability': 0.23}
        }
    }
}

Model Management

# List all available trained models
models = model.list_models()

# Get detailed information about a specific model
model_info = model.get_model_info(model_id)

# Delete a specific model
success = model.delete_model(model_id)

Performance Analysis

# Get performance metrics for all steps or a specific step
metrics = model.get_performance_metrics(step=None)

# Get detailed information about a specific step model
step_info = model.get_step_model_info(step)

# Get dataset statistics
stats = model.describe_datasets()

Dataset Statistics Format

The describe_datasets() method returns statistics for both development and production datasets:

{
    'development': {
        'num_runs': int,          # Total number of runs
        'num_steps': int,         # Number of steps per run
        'error_free_runs': int,   # Number of runs without errors
        'error_runs': int,        # Number of runs with errors
        'training_rows': int,     # Number of training data points
        'unique_runs': int,       # Number of unique runs
        'error_free_pairs': int,  # Number of error-free pairs
        'error_pairs': int        # Number of error pairs
    },
    'production': {
        'num_runs': int,          # Total number of runs
        'num_steps': int          # Number of steps per run
    }
}

Model Serialization

# Convert model state to JSON
json_str = model.to_json()

# Create model instance from JSON
model = AerylModel.from_json(json_str)

Example Usage

Here's a complete example showing how to train a model and save the results:

from aeryl_sdk import AerylModel
import json
import os
from datetime import datetime

def main():
    # Initialize the model
    model = AerylModel(
        dev_path='data/company_research_dev.csv',
        prod_path='data/company_research_prod.csv',
        models_dir='models'
    )
    
    # Train the model
    model_id = model.train()
    
    # Get dataset statistics and performance metrics
    dataset_stats = model.describe_datasets()
    performance_metrics = model.get_performance_metrics()
    predictions = model.predict()
    
    # Create output dictionary
    output = {
        'timestamp': datetime.now().isoformat(),
        'model_id': model_id,
        'dataset_statistics': dataset_stats,
        'performance_metrics': performance_metrics,
        'predictions': predictions
    }
    
    # Save results to JSON file
    os.makedirs('output', exist_ok=True)
    output_file = f'output/model_run_{datetime.now().strftime("%Y%m%d_%H%M%S")}.json'
    with open(output_file, 'w') as f:
        json.dump(output, f, indent=2)

if __name__ == '__main__':
    main()

Requirements

• Python 3.7+
• XGBoost
• NumPy
• Polars
• scikit-learn

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome! Please read our Contributing Guidelines for details on our code of conduct and the process for submitting pull requests.

Contact

For any questions or concerns, please contact us at [e-mail].