aponyx 0.1.1

Systematic research framework for tactical credit strategies

Aponyx

A modular Python framework for developing and backtesting systematic credit strategies.

⚠️ Bloomberg Terminal Required: Data loading via BloombergSource requires an active Bloomberg Terminal session. File-based alternatives (FileSource) are also supported.

Quick Start

Installation from PyPI

pip install aponyx

# Optional: visualization dependencies
pip install aponyx[viz]

Installation from Source

Requires Python 3.12 and uv for environment management.

# Clone repository
git clone https://github.com/stabilefrisur/aponyx.git
cd aponyx

# Create virtual environment with uv
uv venv
source .venv/Scripts/activate  # Windows: .venv\Scripts\activate

# Install dependencies
uv sync

# Install visualization dependencies (optional)
uv sync --extra viz

# Run examples
uv run python examples/backtest_demo.py

Basic Usage

from aponyx.data import fetch_cdx, fetch_etf, FileSource, BloombergSource
from aponyx.models import compute_cdx_etf_basis, SignalConfig
from aponyx.backtest import run_backtest, BacktestConfig

# Load market data with validation (file-based)
cdx_df = fetch_cdx(FileSource("data/raw/cdx_data.parquet"), index_name="CDX_IG_5Y")
etf_df = fetch_etf(FileSource("data/raw/etf_data.parquet"), ticker="HYG")

# Or use Bloomberg Terminal (requires active session and xbbg package)
# source = BloombergSource()
# cdx_df = fetch_cdx(source, index_name="CDX_IG", tenor="5Y")
# etf_df = fetch_etf(source, ticker="HYG")

# Generate signal
signal_config = SignalConfig(lookback=20, min_periods=10)
signal = compute_cdx_etf_basis(cdx_df, etf_df, signal_config)

# Run backtest
backtest_config = BacktestConfig(entry_threshold=1.5, exit_threshold=0.75)
results = run_backtest(signal, cdx_df["spread"], backtest_config)
print(f"Sharpe Ratio: {results.metrics['sharpe_ratio']:.2f}")

Project Structure

aponyx/
├── src/aponyx/       # Core framework
│   ├── data/              # Data loading, validation, transformation
│   ├── models/            # Signal generation for credit strategies
│   ├── backtest/          # Backtesting engine and metrics
│   ├── visualization/     # Plotly charts and Streamlit dashboard
│   ├── persistence/       # Parquet/JSON I/O and registry
│   └── config/            # Configuration and constants
├── examples/              # Runnable demonstrations
├── tests/                 # Unit tests
└── docs/                  # Architecture and design docs

Key Layers

Layer	Purpose	Entry Point
Data	Load, validate, transform market data	aponyx.data
Models	Generate signals for independent evaluation	aponyx.models
Backtest	Simulate execution and compute metrics	aponyx.backtest
Visualization	Interactive charts and dashboards	aponyx.visualization
Persistence	Save/load data with metadata registry	aponyx.persistence

Documentation

Getting Started

• Python Guidelines - Code standards and best practices
• Examples - Runnable demos for each layer

Strategy & Architecture

• CDX Overlay Strategy - Investment thesis and pilot implementation
• Documentation Structure - Single source of truth principles

Design Guides

• Logging Design - Logging conventions and metadata tracking
• Signal Registry Usage - Signal management and research workflow
• Visualization Design - Chart architecture and integration patterns
• Caching Design - Cache layer architecture and usage
• Adding Data Providers - Provider extension guide

Features

✅ Type-safe data loading with schema validation
✅ Modular signal framework with composable transformations
✅ Deterministic backtesting with transaction cost modeling
✅ Interactive visualization with Plotly charts (equity, signals, drawdown)
✅ Parquet-based persistence with JSON metadata registry
✅ Comprehensive logging with run metadata tracking
✅ Bloomberg Terminal integration via xbbg wrapper (included by default, requires active Terminal session)

Planned Features:

• 🔜 Streamlit dashboard (stub exists, not yet implemented)
• 🔜 Advanced attribution charts (stubs with NotImplementedError)

Development

Running Tests

# All tests
uv run pytest

# With coverage
uv run pytest --cov=aponyx --cov-report=term-missing

# Specific module
uv run pytest tests/models/

Code Quality

# Format code
uv run black src/ tests/

# Lint
uv run ruff check src/ tests/

# Type check
uv run mypy src/

Examples

Each example demonstrates a specific layer with synthetic data:

uv run python examples/data_demo.py          # Data loading and transformation
uv run python examples/models_demo.py        # Signal generation
uv run python examples/backtest_demo.py      # Complete backtest workflow
uv run python examples/visualization_demo.py # Interactive charts (requires viz extra)
uv run python examples/persistence_demo.py   # Data I/O and registry

Design Principles

1. Modularity - Separate data, models, backtest, and persistence layers
2. Reproducibility - Deterministic outputs with seed control and metadata logging
3. Type safety - Strict type hints and runtime validation
4. Simplicity - Prefer functions over classes, explicit over implicit
5. Transparency - Clear separation between strategy logic and infrastructure

Architecture Notes

Signal Convention

All model signals follow a consistent sign convention:

• Positive values → Long credit risk (buy CDX / sell protection)
• Negative values → Short credit risk (sell CDX / buy protection)

This convention ensures clear interpretation when evaluating signals individually or when combining signals in future experiments.

Data Flow

Raw Data (Parquet/CSV)
    ↓
Data Layer (load, validate, transform)
    ↓
Models Layer (signal computation)
    ↓
Backtest Layer (simulation, metrics)
    ↓
Persistence Layer (save results, metadata)

License

MIT License - see LICENSE for details.

Contributing

This is a research project under active development. See Python Guidelines for code standards.

---

Maintained by stabilefrisur
Version 0.1.0 | Last Updated: October 31, 2025