Swing Agent v1.6.1

Adds Copilot-friendly instructions.md and scripts/backtest_generate_signals.py (walk-forward historical signal generation). Based on v1.6 enrichments:

• Multi-timeframe alignment (15m + 1h)
• Relative strength vs sector ETF (default XLK) + SPY fallback
• Time-of-day bucket (open/mid/close)
• Volatility regime filter for KNN priors
• Signals DB stores expectations + LLM plan + enrichments
• Centralized database architecture with SQLAlchemy ORM
• External database support (PostgreSQL, MySQL, etc.)

Quick Start for New Users

1. Installation

python -m venv .venv && source .venv/bin/activate
pip install -e .

2. Your First Signal

# Generate a trading signal for Apple stock
python scripts/run_swing_agent.py --symbol AAPL --interval 30m --lookback-days 30

3. Learn the System

• New to SwingAgent? Start with the Getting Started Guide
• Want hands-on experience? Follow the Tutorial
• Have questions? Check the FAQ

Documentation

📚 Complete Documentation - Everything you need to know

For Users

• Getting Started - Setup and first signal
• Tutorial - Your first week of trading
• Use Cases - Real-world trading scenarios
• Best Practices - Professional tips
• FAQ - Common questions
• Glossary - Trading terms explained

For Developers

• API Reference - Function documentation
• Architecture - System design
• Configuration - Customization options
• Development - Contributing guide

What SwingAgent Does

SwingAgent combines technical analysis, machine learning, and AI to generate 1-2 day swing trading signals:

✅ Technical Analysis: Fibonacci retracements, trend analysis, momentum indicators
✅ Pattern Recognition: ML-based historical pattern matching
✅ AI Insights: OpenAI-powered explanations and action plans
✅ Risk Management: Automatic stop-loss and take-profit calculations

Example Signal Output

{
  "symbol": "AAPL",
  "entry": {
    "side": "long",
    "entry_price": 185.50,
    "stop_price": 182.20,
    "take_profit": 190.80,
    "r_multiple": 1.61
  },
  "confidence": 0.72,
  "expected_winrate": 0.58,
  "action_plan": "Strong uptrend with Fibonacci support..."
}

Database Configuration

SwingAgent v1.6.1 uses a centralized database architecture with SQLAlchemy ORM. By default, it uses a single SQLite file, but supports external databases for production deployments.

Default (SQLite)

# Uses data/swing_agent.sqlite automatically
python scripts/run_swing_agent.py --symbol AMD --interval 30m --lookback-days 30

External Database (PostgreSQL/MySQL)

# Option 1: Direct URL
export SWING_DATABASE_URL="postgresql://user:pass@localhost:5432/swing_agent"

# Option 2: Individual components
export SWING_DB_TYPE=postgresql
export SWING_DB_HOST=localhost
export SWING_DB_NAME=swing_agent
export SWING_DB_USER=your_username
export SWING_DB_PASSWORD=your_password

# Then run normally
python scripts/run_swing_agent.py --symbol AMD

CNPG (CloudNativePG) for Kubernetes

# For Kubernetes deployments with CNPG operator
export SWING_DB_TYPE=cnpg
export CNPG_CLUSTER_NAME=swing-postgres
export CNPG_NAMESPACE=default
export SWING_DB_NAME=swing_agent
export SWING_DB_USER=swing_user
export SWING_DB_PASSWORD=your_password

# Test configuration
python scripts/test_cnpg.py

Features

• Centralized Storage: Signals and vector patterns stored in single database
• Multiple Backends: SQLite for development, PostgreSQL/MySQL for production
• Kubernetes Ready: Native CloudNativePG (CNPG) support for scalable deployments
• Migration Tools: Automated migration from legacy separate databases

For detailed database setup:

• EXTERNAL_DATABASES.md - PostgreSQL/MySQL setup
• CNPG_SETUP.md - CloudNativePG for Kubernetes
• k8s/cnpg/ - Complete Kubernetes deployment manifests

Run live signal

python scripts/run_swing_agent.py --symbol AMD --interval 30m --lookback-days 30 --sector XLK

Generate historical signals (no look-ahead)

python scripts/backtest_generate_signals.py --symbol AMD --interval 30m --lookback-days 180 --warmup-bars 80 --sector XLK --no-llm

Evaluate stored signals

python scripts/eval_signals.py --max-hold-days 2.0

Backfill vector store from signal history

python scripts/backfill_vector_store.py

Performance snapshot

python scripts/analyze_performance.py

Train ML models

python scripts/train_ml_model.py --db data/swing_agent.sqlite

Generates baseline classification and regression models from the vector store and saves them to models/.

Database Migration

From Separate Files to Centralized Database

# Migrate existing separate SQLite files to centralized database
python -m swing_agent.migrate --data-dir data/

From SQLite to External Database

# First set up external database connection
export SWING_DATABASE_URL="postgresql://user:pass@localhost:5432/swing_agent"

# Migrate from centralized SQLite to external database
python -m swing_agent.migrate --sqlite-to-external "$SWING_DATABASE_URL"

Database Utilities

# Show current database configuration
python scripts/db_info.py --info

# Test database connection
python scripts/db_info.py --test

# Initialize database tables
python scripts/db_info.py --init