v0.3.4

PySAD v0.3.4 Release

🐛 Critical Bug Fixes

ReferenceWindowModel Reset Bug (Issue #25)

• Fixed critical reference window reset bug that was causing incorrect window reinitialization during sliding window operations
• The bug occurred when cur_window_X length was less than window_size after sliding, causing reference_window_X to be incorrectly reset to just the current window instead of maintaining the properly sized reference window
• This fix ensures stable and reliable sliding window behavior for all models using ReferenceWindowModel (including IForestASD)
• Added comprehensive test coverage to prevent regression

AUROCMetric Single Class Validation

• Added validation for single class presence in AUROCMetric evaluation to prevent evaluation errors
• Now properly handles edge cases where only one class is present in the evaluation data
• Throws informative ValueError with message "Only one class present" instead of cryptic sklearn errors

StreamhashProjector Type Casting Fix

• Fixed hash seed type casting issue in StreamhashProjector's _hash_string method
• Ensures proper integer casting of seed parameter to prevent type-related errors in hash computation
• Improves stability of streamhash projection operations

🔧 Infrastructure Improvements

• Enhanced CI infrastructure with Python 3.13 support
• Updated dependency management for better compatibility
• Improved test coverage for edge cases and bug prevention

🙏 Acknowledgments

Special thanks to @onixlas for discovering and reporting the critical reference window bug that led to this important fix.

📋 Full Changelog

Full Changelog: v0.3.3...v0.3.4

For detailed changes, see our CHANGES.txt.

🚀 Installation

pip install pysad==0.3.4

🔗 Links

• Documentation
• PyPI Package
• Examples

This release focuses on critical stability fixes, particularly for streaming anomaly detection workflows using reference window models. We strongly recommend upgrading from previous versions to ensure reliable sliding window operations.