Skip to content

Signalflow-trading - High-Performance Algorithmic Trading Framework

Current stable version: 0.6.0

signalflow-trading is a high-performance Python framework for algorithmic trading, quantitative finance, and ML-powered signal validation.

What's New in v0.6

  • Fluent Builder API

    Clean, chainable API for backtesting with IDE autocomplete

    result = (
    sf.Backtest("my_strategy")
    .data(raw=my_data)
    .detector("sma_cross")
    .exit(tp=0.03, sl=0.015)
    .run()
)

  • Flow Builder

    Execute strategies as directed flows with metric nodes, LTTB downsampling, and artifact caching

    flow = (
    FlowBuilder("research")
    .data(store="binance", pair="BTC/USDT")
    .detector("sma_cross")
    .metric("sharpe_ratio")
    .build()
)

  • Semantic Decorators

    Type-safe component registration: @sf.detector(), @sf.feature(), @sf.entry(), @sf.exit(), @sf.executor()

  • State Persistence

    StateManager with Redis, DuckDB, and Memory backends for crash recovery and session continuity

  • Statistical Analysis

    Numba-accelerated Monte Carlo, Bootstrap CI, and significance tests

  • CLI & YAML Config

    Run backtests from command line with YAML configuration and search_space() for tuning

  • High Performance

    Polars-first processing for 100+ pairs, 500k+ candles. Numba JIT for statistics

  • Modular Design

    Component registry with pluggable detectors, validators, features, and strategies

  • Production Ready

    Same code for backtesting, paper trading, and live execution

  • ML-Powered

    scikit-learn, XGBoost, LightGBM, PyTorch signal validation with meta-labeling

The Signal Pipeline

flowchart LR
    A[Market Data] --> B[Signal Detection]
    B --> C[Signal Validation]
    C --> D[Strategy Execution]

    style A fill:#14b8a6,stroke:#0d9488,stroke-width:2px,color:#fff
    style B fill:#8b5cf6,stroke:#7c3aed,stroke-width:2px,color:#fff
    style C fill:#3b82f6,stroke:#2563eb,stroke-width:2px,color:#fff
    style D fill:#22c55e,stroke:#16a34a,stroke-width:2px,color:#fff

1. Data & Features

  • 7 Exchange Sources: Binance, Bybit, OKX, Deribit, Kraken, Hyperliquid, WhiteBIT
  • DuckDB, SQLite, PostgreSQL storage backends
  • 189+ technical indicators via signalflow-ta

2. Signal Detection

  • Classical algorithms (SMA crossover, MACD, RSI, structure detection)
  • Neural network predictions via signalflow-nn
  • Custom detectors via @sf.detector("name")

3. Signal Validation (Meta-Labeling)

  • Lopez de Prado's meta-labeling methodology
  • scikit-learn, XGBoost, LightGBM classifiers
  • Deep learning validators via PyTorch Lightning

4. Strategy Execution

  • Entry/exit rules: signal-based, model-driven, composite
  • Position sizing: Kelly Criterion, volatility targeting, risk parity, martingale
  • Entry filters: regime, volatility, drawdown, correlation, time-of-day
  • Signal aggregation: majority, weighted, unanimous, meta-labeling
  • Real-time: Paper trading with RealtimeRunner, monitoring, alerts
  • State persistence: Redis/DuckDB/Memory with crash recovery

Quick Example

import signalflow as sf

result = (
    sf.Backtest("momentum_strategy")
    .data(raw=my_raw_data)
    .detector("example/sma_cross", fast_period=20, slow_period=50)
    .entry(size_pct=0.1, max_positions=5)
    .exit(tp=0.03, sl=0.015, trailing=0.02)
    .capital(50_000)
    .fee(0.001)
    .run()
)

print(result.summary())
result.plot()

Custom Components

import signalflow as sf
from signalflow.detector import SignalDetector

@sf.detector("my/custom_detector")
class CustomDetector(SignalDetector):
    def detect(self, data):
        return signals

Available decorators: @sf.detector(), @sf.feature(), @sf.entry(), @sf.exit(), @sf.executor(), @sf.data_source(), @sf.data_store(), @sf.strategy_store(), @sf.register()

Key Features

Polars-First Performance

Core data processing uses Polars for extreme efficiency on large datasets, with seamless Pandas compatibility.

Advanced Labeling

Triple Barrier Method, Fixed Horizon, Trend Scanning, Volatility and Volume labelers. Numba-accelerated (45s to 0.3s).

189+ Technical Indicators

The signalflow-ta extension: momentum, volatility, trend, statistics, and physics-based market analogs.

Paper Trading & Monitoring

RealtimeRunner with async data sync, virtual broker, drawdown/stuck-position alerts, and crash recovery.

Statistical Validation

Monte Carlo simulation, Bootstrap confidence intervals, significance tests — all Numba JIT-compiled.

Flow Builder

Execute strategies as directed acyclic graphs with metric nodes, progress callbacks, and LTTB downsampling.

Technology Stack

  • Polars — High-performance DataFrames
  • DuckDB — Embedded analytics database
  • Pandas — Compatibility layer
  • NumPy — Numerical computing
  • scikit-learn — Classical ML models
  • XGBoost / LightGBM — Gradient boosting
  • PyTorch + Lightning — Deep learning
  • Optuna — Hyperparameter optimization
  • Numba — JIT compilation
  • signalflow-ta — 189+ indicators
  • pandas-ta — TA foundation
  • Plotly — Interactive charts
  • Redis — State persistence (production)
  • DuckDB / SQLite / PostgreSQL — Storage
  • Kedro — Pipeline orchestration
  • FastAPI — Web backend (sf-ui)

SignalFlow Ecosystem

  • signalflow-trading (Core)

    Data containers, signal detection, backtesting, strategy execution, state persistence, statistical analysis

    pip install signalflow-trading

  • signalflow-ta

    189+ technical indicators, 24 signal detectors, physics-based market analogs, AutoFeatureNormalizer

    pip install signalflow-ta

  • signalflow-nn

    14 neural encoders (LSTM, Transformer, PatchTST, TCN), 7 heads, 4 loss functions

    pip install signalflow-nn

  • sf-kedro

    Universal ML pipelines: backtest, analyze, train, tune, validate with Optuna and MLflow

Getting Started

Support

License

SignalFlow is open source software released under the MIT License.

Disclaimer

SignalFlow is provided for research purposes. Trading financial instruments carries risk. Past performance does not guarantee future results. Use at your own risk.