Release v0.1.10 - Raw Data Storage and Performance Enhancements · stabilefrisur/aponyx

Added

Comprehensive raw data storage system with hash-based file naming across all providers
- Unified storage under data/raw/synthetic/, data/raw/file/, data/raw/bloomberg/
- Deterministic hash-based naming for deduplication and integrity
- Metadata JSON files tracking fetch parameters and timestamps
- Test suite validating storage patterns across providers
Intraday data update capability in single-signal notebook template
- update_cached_data() workflow for Bloomberg refresh
- Configurable field subsets and refresh intervals
- Automatic cache validation and staleness checks
Signal lag parameter sweep example demonstrating systematic optimization
- Grid search across entry/exit thresholds and signal lags
- Heatmap visualization of performance surfaces
- Best parameter identification workflow
Sign multiplier feature in signal catalog for flexible signal inversion
- sign_multiplier field in signal catalog JSON (+1 or -1)
- Automatic signal inversion without code changes
- Enhanced signal registry with multiplier validation

Refactored data architecture to proper raw/cache/processed separation
- Raw storage: immutable source data with metadata
- Cache: TTL-based Parquet for fetch optimization
- Processed: analysis-ready transformed datasets
- Clear separation of concerns and data lineage
Enhanced synthetic data generation with improved market dynamics
- More realistic correlation structures between CDX/VIX/ETF
- Configurable regime simulation (normal, stress, recovery)
- Better alignment with actual market behavior patterns
Streamlined example scripts with improved Bloomberg integration
- Enhanced error handling for Terminal connection issues
- Better fallback patterns for offline development
- Clearer documentation of Bloomberg requirements
Integrated quantstats library for enhanced performance analytics
- Extended metrics beyond internal calculations
- Professional tearsheet generation capability
- Comprehensive risk-adjusted performance analysis
Converted backtest adapters to protocol-based interface
- Clean separation between internal engine and third-party libraries
- Extensible adapter pattern for quantstats/vectorbt integration
- Improved testability and modularity

Bloomberg provider now correctly handles BDP/BDH format differences
- BDP returns dict format for static fields
- BDH returns DataFrame format for time series
- Proper handling of both response types
Suitability evaluation import path corrected in Quick Start example
- Fixed module path in README.md demonstration
- Updated to match actual package structure

Added comprehensive raw data storage design document
- Hash-based naming rationale and implementation
- Provider-specific storage patterns
- Metadata schema and validation approach
Updated caching design documentation to reflect architecture changes
- Clarified cache vs raw storage distinction
- Enhanced provider integration patterns
- Improved cache invalidation workflows