┌──────────────┐     ┌──────────────┐
│ candles.db   │────▶│              │
└──────────────┘     │   Signal     │     ┌─────────────────┐
                     │  Generator   │────▶│ Unix Socket     │
┌──────────────┐     │              │     │ /tmp/signals.sock│
│ analysis.db  │────▶│              │     └─────────────────┘
└──────────────┘     └──────────────┘              │
                            │                      ▼
                            │              ┌─────────────────┐
                            │              │ Trading Bot     │
                            ▼              └─────────────────┘
                     ┌─────────────────┐
                     │ Health Socket   │
                     │ /tmp/signals_   │
                     │ health.sock     │
                     └─────────────────┘

Installation

1. Install TA-Lib System Library

Ubuntu/Debian:

wget http://prdownloads.sourceforge.net/ta-lib/ta-lib-0.4.0-src.tar.gz
tar -xzf ta-lib-0.4.0-src.tar.gz
cd ta-lib/
./configure --prefix=/usr
make
sudo make install

macOS:

brew install ta-lib

2. Install Python Dependencies

pip install -r requirements.txt

Configuration

Edit config.json to customize behavior:

Key Parameters

Parameter	Description	Default
`personality`	Trading style: "scalping" or "swing"	"scalping"
`timeframes`	List of timeframes to monitor	["1m", "5m"]
`poll_interval`	Seconds between checks	0.5
`min_confidence`	Minimum confidence threshold (0-1)	0.6
`cooldown_seconds`	Cooldown between signals	60
`lookback`	Candles to fetch for analysis	200

Personality Configurations

Scalping (Short-term, high-frequency):

Timeframes: 1m, 5m
Focus: EMA crossovers, Stochastic, volume surges
Best for: Quick moves, scalp trades
Signal frequency: Higher (10-50/day)

Swing (Medium-term, trend-following):

Timeframes: 15m, 1h
Focus: Regime filters, Bollinger squeezes, MACD
Best for: Trend captures, larger moves
Signal frequency: Lower (2-10/week)

Signal Weights

Adjust weights in config.json to tune signal generation:

"weights": {
  "scalping": {
    "ema_cross": 0.3,    // EMA 9/21 crossover
    "stoch": 0.25,       // Stochastic oscillator
    "rsi": 0.2,          // RSI oversold/overbought
    "volume": 0.15,      // Volume surge
    "macd": 0.1          // MACD confirmation
  },
  "swing": {
    "regime": 0.35,      // SMA 50/200 filter
    "bb_squeeze": 0.25,  // Bollinger squeeze
    "macd": 0.2,         // MACD crossover
    "flow": 0.15,        // Buy/sell pressure
    "rsi": 0.05          // RSI filter
  }
}

Running

Start the Generator

python signal_generator.py

Or with custom config:

python signal_generator.py -c custom_config.json

Run as Background Service

nohup python signal_generator.py > /dev/null 2>&1 &

Systemd Service (Recommended)

Create /etc/systemd/system/signal-generator.service:

[Unit]
Description=BTCUSDT Signal Generator
After=network.target

[Service]
Type=simple
User=youruser
WorkingDirectory=/path/to/bybitbtc/signals
ExecStart=/usr/bin/python3 signal_generator.py
Restart=always
RestartSec=10

[Install]
WantedBy=multi-user.target

Then:

sudo systemctl daemon-reload
sudo systemctl enable signal-generator
sudo systemctl start signal-generator
sudo systemctl status signal-generator

Health Checks

Query Current Status

python health_check.py

Output example:

============================================================
SIGNAL GENERATOR HEALTH STATUS
============================================================
Status:           running
Personality:      scalping
Timeframes:       1m, 5m
Uptime:           3456s (57m)
Total Signals:    23
  - Buy:          14
  - Sell:         9
Last Signal:      2026-01-15T14:32:11.234Z
Errors:           0
Connected Clients: 1

Recent Signals:
  [1m] BUY @ $43250.50 (conf: 0.75)
      Reasons: EMA9 crossed above EMA21, Stochastic oversold crossover
  [5m] SELL @ $43180.25 (conf: 0.68)
      Reasons: EMA9 crossed below EMA21, RSI oversold (62.3)
============================================================

Monitor Logs

tail -f logs/signal_generator.log

Testing the Signal Stream

Connect Test Client

python test_client.py

This will connect and display all incoming signals in real-time:

======================================================================
[2026-01-15 14:32:11] BUY SIGNAL
======================================================================
Timeframe:   1m
Price:       $43250.50
Confidence:  75.0%
Personality: scalping
Reasons:
  • EMA9 crossed above EMA21
  • Stochastic oversold crossover
  • Volume surge detected
======================================================================

Integration with Trading Bot

Your trading bot should connect to /tmp/signals.sock:

import socket
import json

sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
sock.connect("/tmp/signals.sock")

buffer = ""
while True:
    chunk = sock.recv(4096).decode('utf-8')
    buffer += chunk
    
    while '\n' in buffer:
        line, buffer = buffer.split('\n', 1)
        if line.strip():
            signal = json.loads(line)
            # Process signal
            print(f"Received: {signal['signal']} @ ${signal['price']}")

Signal Format

Each signal is a JSON object with the following structure:

{
  "signal": "BUY",
  "timeframe": "1m",
  "confidence": 0.75,
  "price": 43250.50,
  "timestamp": 1705329131,
  "reasons": [
    "EMA9 crossed above EMA21",
    "Stochastic oversold crossover",
    "Volume surge detected"
  ],
  "personality": "scalping",
  "generated_at": "2026-01-15T14:32:11.234567Z"
}

Signal Logic

Scalping Strategy

Generates signals on:

EMA Crossover: 9/21 cross with direction confirmation
Stochastic: Oversold (<30) or overbought (>70) with crossover
RSI: Confirmation filter (<40 buy, >60 sell)
Volume: 1.5x average indicates momentum
MACD: Directional confirmation

Swing Strategy

Generates signals on:

Regime Filter: Price vs SMA 50/200 trend structure
BB Squeeze: Volatility compression + breakout
MACD: Crossover for momentum shift
Flow: Buy/sell pressure ratio (55% threshold)
RSI: Light filter to avoid extremes

Performance Tuning

Reduce Signal Frequency

Increase min_confidence:

"min_confidence": 0.75

Increase cooldown_seconds:

"cooldown_seconds": 300

Increase Sensitivity

Lower confidence threshold:

"min_confidence": 0.5

Reduce cooldown:

"cooldown_seconds": 30

Change Polling Rate

Faster updates (higher CPU):

"poll_interval": 0.2

Slower updates (lower CPU):

"poll_interval": 1.0

Troubleshooting

"Socket not found" Error

Check if generator is running:

python health_check.py

Check for socket file:

ls -la /tmp/signals*.sock

No Signals Generated

Check database connections:

sqlite3 ../onramp/candles.db "SELECT COUNT(*) FROM candles WHERE timeframe='1m';"

Lower confidence threshold temporarily:
```
"min_confidence": 0.4
```
Check logs for errors:
```
grep ERROR logs/signal_generator.log
```

Database Locked Errors

Ensure WAL mode is enabled (automatic in code):

sqlite3 ../onramp/candles.db "PRAGMA journal_mode=WAL;"

High CPU Usage

Increase poll interval:

"poll_interval": 1.0

Reduce lookback:

"lookback": 100

Maintenance

Log Rotation

Use logrotate or manually clear:

echo "" > logs/signal_generator.log

Database Optimization

The generator reads in WAL mode and doesn't modify data, so no maintenance needed on its side.

Next Steps

Backtest: Use historical data to validate signal quality
Tune Weights: Adjust based on paper trading results
Add Personalities: Create custom strategies (e.g., "momentum", "mean-reversion")
ML Integration: Replace rule-based logic with trained models
Multi-Asset: Extend to other trading pairs

Notes

Signals are based on closed candles only - no repainting risk
The latest forming candle is automatically excluded
All timestamps are UTC
Confidence scores are additive from multiple indicators
Cooldown prevents signal spam per timeframe/personality combination