TradeAgent is a production-ready hybrid AI trading system that combines deterministic technical analysis with LLM-powered sentiment analysis. Built for Alpaca's paper trading API, it features adaptive parameter optimization, portfolio correlation monitoring, and comprehensive risk management.

• Momentum Trading: RSI, MACD, volume-based signals with dynamic thresholds
• News Sentiment Analysis: Claude 3.5 Sonnet analyzes 90+ articles per ticker
• Sentiment Trend Tracker: Detects sentiment momentum shifts and inflection points
• Defensive Core: Monthly rebalancing of VTI/VGK/GLD allocation

• Adaptive Parameter Optimizer: Grid search optimization based on rolling 30-day performance
• ML Training Pipeline: Collects labeled data for future ensemble models
• Performance Analytics: Daily & weekly reports with Sharpe ratio tracking

• Portfolio Correlation Monitor: Blocks over-concentrated positions (max 40% per sector)
• Dynamic Position Sizing: Kelly Criterion with half-Kelly safety (3-15% per position)
• Circuit Breakers: -3% daily loss limit, automatic trading halt
• Stop-Loss & Take-Profit: Bracket orders on all momentum entries

• Multi-Source News: Yahoo Finance, Finnhub, Alpha Vantage, NewsAPI
• Persistent Logging: All trades, signals, and LLM analyses stored in Supabase
• Complete Audit Trail: Track parameter changes, performance metrics, reasoning

• Python 3.12+
• Alpaca Paper Trading Account (Sign up)
• Supabase Account (Sign up)
• OpenRouter API Key (Sign up) (optional, for LLM features)

1. Clone the repository ```bash git clone https://github.com/enving/TradeAgent.git cd TradeAgent ```

2. Create virtual environment ```bash python3 -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate ```

3. Install dependencies ```bash pip install -r requirements.txt ```

4. Configure environment variables ```bash cp .env.example .env

```

Required variables: ```env ALPACA_API_KEY=your_alpaca_key ALPACA_SECRET_KEY=your_alpaca_secret SUPABASE_URL=your_supabase_url SUPABASE_KEY=your_supabase_anon_key ```

Optional (for LLM features): ```env OPENROUTER_API_KEY=your_openrouter_key NEWS_API_KEY=your_newsapi_key FINNHUB_API_KEY=your_finnhub_key ENABLE_LLM_FEATURES=true ```

5. Set up Supabase database

Run these SQL migrations in your Supabase SQL Editor (in order):

1. database/migrations/create_ml_training_data.sql

2. database/migrations/add_news_and_llm_logging_v2.sql

3. database/migrations/add_parameter_changes.sql

4. Test the installation ```bash python3 test_integration.py ```

Run the bot autonomously with automated scheduling!

```bash

./run_docker.sh build

./run_docker.sh once

./run_docker.sh schedule

./run_docker.sh logs

./run_docker.sh stop ```

• ✅ Autonomous operation - Runs daily at 9:35 AM ET automatically
• ✅ Isolated environment - No dependency conflicts
• ✅ Auto-restart - Recovers from failures automatically
• ✅ Cross-platform - Works on Linux, macOS, Windows
• ✅ Easy updates - `git pull && ./run_docker.sh build`

For native Linux deployment or cron-based scheduling, see detailed guide:

• deployment/README.md - Full deployment guide with SystemD and Cron options

```bash python3 -m src.main ```

The bot will:

1. Check market hours
2. Scan for momentum signals (15 tickers)
3. Analyze news sentiment (if LLM enabled)
4. Apply risk filters (correlation, sector limits)
5. Execute trades with bracket orders
6. Monitor exit conditions for open positions
7. Log performance metrics to Supabase

```bash python3 -c "import asyncio; from src.main import daily_trading_loop; asyncio.run(daily_trading_loop(allow_premarket=True))" ```

After collecting 30+ trades: ```bash

python3 run_optimizer.py --strategy momentum --lookback 30

python3 run_optimizer.py --show ```

```bash

python3 test_correlation_monitor.py

python3 test_sentiment_tracker.py ```

``` TradeAgent/ ├── src/ │ ├── strategies/ # Trading strategies │ ├── risk/ # Risk management │ ├── ml/ # Machine learning │ ├── llm/ # LLM integration │ ├── core/ # Core logic │ ├── adapters/ # External APIs │ ├── models/ # Data models │ └── database/ # Supabase client ├── database/migrations/ # SQL migrations └── tests/ # Tests ```

Entry: RSI 45-75, MACD > 0, Volume > 1.1x avg
Exit: Stop -3%, Target +8%, RSI > 80

Entry: Sentiment > 0.6, High impact, OR inflection detected
Claude 3.5 Sonnet analyzes 90+ articles/ticker

Allocation: 50% VTI, 30% VGK, 20% GLD
Rebalance: Monthly, first trading day

Tests 27+ parameter combinations on 30-day performance, selects optimal Sharpe ratio.

Detects sentiment momentum via linear regression, blocks volatile signals.

Prevents over-concentration: max 0.7 correlation, 40% sector limit.

```bash

python3 test_integration.py

python3 test_correlation_monitor.py python3 test_sentiment_tracker.py ```

Educational purposes only. Not financial advice.

• Paper trading only
• No profitability guarantees
• Trading involves risk of loss
• Use at your own risk

MIT License - see LICENSE

Built with ❤️ by algorithmic traders, for algorithmic traders.