Roadmap¶
Project roadmap and planned features for shap-monitor.
Current Status¶
Version: 0.0.1 (Work in Progress)
shap-monitor is in early development. The core functionality is being actively developed.
Version History¶
v0.0.1 (Current)¶
Core functionality:
- SHAPMonitor for logging SHAP explanations
- ParquetBackend for efficient storage
- SHAPAnalyzer for basic analysis
- Summary statistics over time periods
- Time period comparison
- Support for tree-based explainers
Planned Features¶
v0.2 (Planned)¶
Drift Detection
- Automatic drift detection in feature importance
- Configurable drift thresholds
- Alerting capabilities
Asynchronous Processing
- Async logging for production systems
- Background workers for SHAP computation
- Queue-based architecture
MLflow Integration
- Log SHAP values to MLflow experiments
- Compare explanations across MLflow runs
- Integration with MLflow Model Registry
Model Version Comparison
- Compare explanations across model versions
- A/B testing support
- Champion/challenger analysis
v0.3+ (Future)¶
Dashboard & Visualization
- Web-based dashboard for monitoring
- Interactive visualizations
- Real-time monitoring
Additional Integrations
- Weights & Biases integration
- Neptune.ai integration
- Custom integration framework
Advanced Features
- Explanation clustering
- Anomaly detection in explanations
- Explanation stability metrics
- Custom drift metrics
Additional Backends
- S3/Cloud storage support
- Database backends (PostgreSQL, DuckDB)
- Time-series database integration
Design Principles¶
- Production-First: Built for production ML systems
- Minimal Overhead: Low latency impact on predictions
- Flexible Storage: Pluggable backend architecture
- Easy Integration: Works with existing ML pipelines
- Comprehensive Analysis: Rich analysis capabilities
Contributing¶
Want to help build these features? See the Contributing Guide for how to get involved.
Feedback¶
Have suggestions for the roadmap? Open a discussion or file an issue.