Microsoft Trading Bot

AI-powered trading bot using TradingAgents + NVIDIA NIM + Twelve Data.

How It Works

Two-phase analysis for full quality at speed:

Phase 1 (5 min):     Quick scan 1000 stocks -> Top 20 candidates
Phase 2 (90 min):    Full TradingAgents deep analysis on candidates

Phase	Method	Time	Quality
1	Single LLM call per stock	~5 min for 1000	Quick filter
2	8+ agents, bull/bear debate	~5 min per stock	Full research

Setup

1. Clone & Install

git clone https://github.com/Rohan5commit/microsoft-trading-bot.git
cd microsoft-trading-bot
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

2. Configure API Keys

cp .env.example .env

Required:

NVIDIA_API_KEY — Free from build.nvidia.com
TWELVE_DATA_KEYS — From twelvedata.com

3. Verify Setup

python -c "from bot.two_phase_bot import TwoPhaseBot; print('Setup OK')"

Usage

Analyze Specific Stocks

python bot/two_phase_scheduler.py --tickers NVDA AAPL MSFT --deep-count 3

Full Universe (1000 stocks)

python bot/two_phase_scheduler.py --deep-count 20

GitHub Actions

Runs daily at 2:00 PM ET (18:00 UTC) on weekdays during US market hours.

Manual Trigger

gh workflow run daily-analysis.yml

# With specific tickers
gh workflow run daily-analysis.yml -f tickers="NVDA,AAPL,MSFT" -f deep_count=5

Secrets Required

Set in GitHub repo Settings > Secrets:

Secret	Required
`NVIDIA_API_KEY`	Yes
`TWELVE_DATA_KEYS`	Yes
`ALPACA_API_KEY`	Optional (for trading)
`ALPACA_SECRET_KEY`	Optional (for trading)
`SENDER_EMAIL`	Optional (for email notifications)
`RECEIVER_EMAIL`	Optional (for email notifications)
`GMAIL_APP_PASSWORD`	Optional (for email notifications)

Files

File	Purpose
`two_phase_bot.py`	Main two-phase analysis engine
`two_phase_scheduler.py`	Entry point for GitHub Actions
`twelve_data.py`	Market data with 8-key rotation
`risk_manager.py`	Position sizing & risk rules
`alpaca_client.py`	Alpaca paper/live trading wrapper
`email_sender.py`	Gmail SMTP daily update emails
`universe.py`	Stock universe management
`portfolio.py`	Portfolio tracking with leverage-adjusted returns
`config.json`	Configuration

Configuration

Edit bot/config.json:

{
  "universe": {
    "max_stocks": 100,
    "min_market_cap_billion": 2
  },
  "deep_analysis": {
    "count": 20,
    "min_conviction": 0.3
  },
  "llm": {
    "provider": "nvidia",
    "deep_think_model": "meta/llama-3.1-70b-instruct",
    "quick_think_model": "meta/llama-3.1-8b-instruct"
  }
}

Risk Warning

This is for educational purposes. Trading involves risk of loss. Start with paper trading.

Rohan5commit/microsoft-trading-bot

README