AI Signal Blueprint

Introduction

System Architecture

Feature Engineering

80% of the system's edge comes from robust feature engineering. RabitAI fuses technical, volatility, and structure features across multiple timeframes.

def extract_features(df):
    features = {}
    # ...calculate indicators...
    features['ema_gap'] = df['ema9'].iloc[-1] - df['ema21'].iloc[-1]
    features['rsi'] = df['rsi'].iloc[-1]
    features['adx'] = df['adx'].iloc[-1]
    features['atr'] = df['atr'].iloc[-1]
    features['cci'] = df['cci'].iloc[-1]
    return features

Market Regime Detection

RabitAI adapts to trending and mean-reverting markets by classifying regimes using entropy, ADX, and volatility relationships.

def detect_regime(features):
    if features['entropy'] < 1.0 and features['adx'] > 25:
        return 'momentum'
    elif features['entropy'] > 1.7 and features['adx'] < 20:
        return 'mean-reverting'
    return 'neutral'

Probability Scoring

Each signal is scored for confidence using a logistic model, trend scaling, entropy filtering, and sentiment alignment. Only high-confidence trades are sent.

def calculate_trade_probability(features):
    base = 0.0
    base += 0.3 * (features['ema_gap'])
    base += 0.25 * (features['rsi'] - 50)/50
    base += 0.2 * (features['adx'] / 50)
    base -= 0.15 * features['entropy']
    prob = 1 / (1 + math.exp(-base))
    return prob * 100

News & Sentiment Filtering

RabitAI integrates news headlines, client sentiment, VADER, and GPT-based summaries to avoid trading against major news flow.

1. Pull latest headlines per ticker
2. Score with VADER & keyword filters
3. Pass to GPT with indicator context
4. Compare GPT’s direction with signal
5. Penalize probability if mismatch

sentiment_score = analyzer.polarity_scores(news_text)['compound']
if sentiment_score < -0.3 and signal == 'BUY':
    confidence *= 0.8

TP/SL Prediction

Take Profit and Stop Loss levels are set dynamically based on volatility and market regime, not static percentages.

• ATR as volatility proxy
• High entropy → tighter SL
• Strong signals → wider TP

def calculate_tp_sl(current_price, atr, regime):
    if regime == 'momentum':
        sl = current_price - atr * 1.2
        tp = current_price + atr * 2.5
    elif regime == 'mean-reverting':
        sl = current_price - atr * 1.0
        tp = current_price + atr * 1.5
    else:
        sl = current_price - atr * 1.1
        tp = current_price + atr * 2.0
    return round(tp, 2), round(sl, 2)

Signal Reasoning with GPT

Every trade is reviewed by GPT-4, which analyzes technicals, sentiment, price structure, and volatility, then provides a rationale.

prompt = f"""
Analyze the following trade setup:
Symbol: {{symbol}}
Direction: {{direction}}
Timeframes: 15m, 30m, 1H, 4H
Indicators: {{indicator_summary}}
News Headlines: {{headline_summary}}
Entropy: {{entropy}}
ATR: {{atr}}

Should this trade be executed? If yes, why? What risk factors are present?
"""

Trade Report Output

Qualified signals are packaged into a polished HTML report, sent via email or Discord. Each report includes:

• Symbol, timeframes, entry, SL, TP
• Direction, confidence score
• GPT-generated summary
• News sentiment notes
• Embedded chart with TP/SL

fig, ax = plt.subplots()
ax.plot(df['close'], label='Price')
ax.axhline(tp, color='green', linestyle='--', label='TP')
ax.axhline(sl, color='red', linestyle='--', label='SL')
# ...
img_buf = io.BytesIO()
plt.savefig(img_buf, format='png')
img_base64 = base64.b64encode(img_buf.getvalue()).decode()

Portfolio Integration & Risk

RabitAI manages risk at the portfolio level, adjusting position size and filtering overlapping or correlated trades.

• Position sizing based on confidence and volatility
• Limits on sector/correlation exposure
• Daily portfolio summary output

def determine_position_size(probability, atr, account_balance):
    risk_pct = min(0.02, probability / 100 * 0.03)
    dollar_risk = account_balance * risk_pct
    position_size = dollar_risk / atr
    return round(position_size, 2)

{
  "date": "2025-05-26",
  "total_signals": 4,
  "executed": 2,
  "capital_allocated": "$4,000",
  "estimated_risk": "$120",
  "expected_return": "$360"
}

Lessons Learned

Join the Mission

RabitAI is now public. Choose your path:

1. Free 30-day trial: Test daily signals in our private Discord
2. Monthly subscription: Real-time alerts, email reports, Discord access
3. Build your own: Use this blueprint, adapt the code, or hire us for custom solutions

Appendix

from dotenv import load_dotenv
load_dotenv('signals_cred.env')
OPENAI_API_KEY = os.getenv('OPENAI_API_KEY')