Multi-Currency Spot Trader (OKX + TA + Optional DeepSeek)

A pragmatic, menu-driven spot trading system for OKX.
It fetches market data, computes robust technical signals, optionally blends an AI opinion, sizes positions under strict risk constraints, and executes market orders with dynamic trailing stops and full state persistence.

Features

• OKX official SDK spot trading (market orders)
• Technical stack: RSI/KDJ/MACD/Bollinger/ATR/MA/Stochastic + reversal detectors
• Decisioning: TA weighted score + (optional) DeepSeek AI JSON action
• Risk control: max portfolio allocations, fixed SL/TP, trailing stops that ratchet up
• Persistence: SQLite for positions, trades, trailing stop state
• Multi-symbol orchestration with a simple CLI control panel

Project layout

apiratelimiter.py      # tiny per-API rate limiter
database_manager.py    # SQLite schema + CRUD for positions/trades/dynamic stops
deepseekanalyzer.py    # optional AI integration (BUY/SELL/HOLD + sell proportion)
multistrategyrunner.py # threads per symbol, lifecycle, status
okxapiclient.py        # OKX SDK wrapper: candles, prices, balances, orders
okxtrading2.0.py       # CLI entrypoint (menu)
riskmanager.py         # portfolio sizing, limits, trailing parameters
technicalanalyzer.py   # indicators, signals, reversal detectors
tradingstrategy.py     # analysis→decision matrix→execution loop per symbol

Requirements

• Python 3.10+
• Packages: okx (official SDK), pandas, numpy, python-dotenv (optional), plus stdlib.
• An OKX API key with trading permissions for spot.

  By default tdMode is set to "cross" and instruments are fetched with instType='SPOT'.

Quick start

1. Install deps

python -m venv .venv && source .venv/bin/activate
pip install okx pandas numpy python-dotenv

2. Set environment variables

Create .env (or export in your shell):

# OKX (required)
OKX_API_KEY=...
OKX_SECRET_KEY=...
OKX_PASSWORD=...

# Optional AI
DEEPSEEK_API_KEY=...

# Risk/sizing (defaults shown)
MAX_TOTAL_POSITION=1          # 100% portfolio cap
MAX_SINGLE_POSITION=0.3       # 30% per symbol cap
STOP_LOSS=0.05                # fixed SL 5%
TAKE_PROFIT=0.15              # fixed TP 15%
TRAILING_STOP_PERCENT=0.03    # 3% trail width
MONITOR_INTERVAL=300          # seconds between trail checks
DB_PATH=trading_system.db     # sqlite path

3. Run the control panel

python okxtrading2.0.py

You’ll see a menu:

• Start All Strategies — launch threads for all configured symbols
• Start/Stop Specific Strategy — control a single market
• View System Status — running flags, active symbols, stored positions
• Manual Position Sync — reconcile on-exchange balances to local state
• Toggle Debug Mode — echo OKX HTTP call codes

How it trades (in one pass)

1. Market data: fetch latest candles for the symbol (1H by default).
2. Indicators: compute RSI/KDJ/MACD/BB/ATR/MA/Stochastic and reversal hints.
3. Signals: produce a weighted TA score and optional DeepSeek action.
4. Decision matrix: combine AI, TA, support/resistance, market state into a total score → BUY/SELL/HOLD + confidence.
5. Sizing & risk: compute size from portfolio constraints and market context (vol/trend/USDT ratio).
6. Execution:
   • BUY: market buy using quote amount (USDT); respects exchange minSz.
   • SELL: market sell using base amount; optional AI estimates a sell proportion (with sensible fallbacks).
7. Stops & persistence:
   • Save position/trades to SQLite.
   • Set/ratchet trailing stop (and a symmetric dynamic take-profit band).
   • Background monitor checks triggers every MONITOR_INTERVAL seconds.

Configuration

• Symbols & cadence: edit symbol_configs in multistrategyrunner.py:

  'ETH-USDT': {'name': 'Ethereum', 'interval': 1800, 'timeframe': '1H'}
  

• Risk knobs: tweak env vars in .env (caps, SL/TP, trailing %).
• Database: change DB_PATH to relocate the SQLite file.

Operational notes

• AI is optional: if DEEPSEEK_API_KEY is missing, system runs purely on TA with deterministic fallbacks.
• Min trade sizes: on buy, system checks if estimated_base_amount >= minSz; on sell, blocks dust/too-small partials.
• Graceful shutdown: Ctrl-C or menu exit triggers strategy/thread stops and monitor shutdown; state is persisted.