Skip to content

Changelog

All notable changes to SignalFlow are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.5.0] - 2026-02-14

Added

  • Fluent Builder API: sf.Backtest() with method chaining for intuitive backtest configuration. Multi-component support with named data sources, detectors, validators, entries, and exits. Signal aggregation modes: merge, majority, weighted, unanimous, any, meta_labeling. Cross-referencing between components (e.g., link entries to specific detectors).

  • One-liner shortcuts: sf.backtest() and sf.load() convenience functions for quick prototyping without full builder configuration.

  • BacktestResult: Rich result container with .summary(), .plot(), .metrics, .metrics_df, and Jupyter HTML rendering via _repr_html_().

  • CLI Interface: Command-line tools for backtest management:

  • sf init -- scaffold a sample YAML configuration
  • sf validate <config.yaml> -- validate configuration without running
  • sf run <config.yaml> -- execute backtest from YAML with optional --plot and -o output

  • sf list detectors|metrics|features|all -- discover available components

  • YAML Configuration: Full backtest configuration via YAML files with component registration, cross-referencing, and validation.

  • Pipeline Visualization: Interactive D3.js HTML visualization of backtest pipelines. Mermaid diagram export for documentation. Local development server via sf.viz.serve(). Graph extractors for BacktestBuilder, FeaturePipeline, and multi-source data flows.

  • OHLCV Resampling: Unified timeframe resampling module (signalflow.data.resample):

  • resample_ohlcv() -- resample OHLCV data between timeframes
  • align_to_timeframe() -- auto-detect source TF and resample
  • detect_timeframe() -- detect timeframe from timestamp intervals
  • select_best_timeframe() -- find best exchange-supported timeframe
  • can_resample() -- check if resampling between two TFs is possible

  • Support for 12 standard timeframes across 8 exchanges

  • Auto-resampling: RawDataFactory.from_stores() and from_duckdb_spot_store() accept target_timeframe parameter for automatic resampling during data loading.

  • Funding Rate Detector: FundingRateDetector for perpetual futures that detects long entry opportunities when funding rate transitions from sustained positive to negative. Registered as "funding/rate_transition".

  • Custom Exception Hierarchy: SignalFlowError base with specialized exceptions (DetectorNotFoundError, MissingDataError, InvalidParameterError, etc.) providing actionable error messages with fix suggestions.

  • Lazy imports: sf.Backtest, sf.BacktestBuilder, sf.BacktestResult, sf.backtest, sf.load, and sf.viz are lazy-loaded to reduce import overhead.

  • New exchange data sources: Deribit, Kraken, Hyperliquid, WhiteBIT (spot & futures), Bybit inverse futures. Total: 7 exchanges with 15 loaders.

  • Multi-source data access: RawData supports hierarchical multi-source access. RawDataLazy container for efficient deferred data loading.

  • Aggregated Open Interest: Features for total OI, OI change, and OI z-score across single and multi-source data.

  • Signal & strategy analytics: Classification, profile, distribution, and result metrics for signals and strategy performance analysis.

Changed

  • Standardized signal type naming and generalized signal handling across detectors and strategy components.

[0.4.1] - Previous release

See GitHub releases for earlier versions.