Quick answers to common questions. For detailed troubleshooting, see Troubleshooting Guide.
NOFX is an AI-powered cryptocurrency trading bot that uses large language models (LLMs) to make trading decisions on futures markets.
- ✅ Binance Futures
- ✅ Hyperliquid
- 🚧 More exchanges coming soon
AI trading is experimental and not guaranteed to be profitable. Always start with small amounts and never invest more than you can afford to lose.
Yes! NOFX supports running multiple traders with different configurations, AI models, and trading strategies.
- OS: Linux, macOS, or Windows (Docker recommended)
- RAM: 2GB minimum, 4GB recommended
- Disk: 1GB for application + logs
- Network: Stable internet connection
No! NOFX has a web UI for all configuration. However, basic command line knowledge helps with setup and troubleshooting.
- Binance: Account → API Management → Create API → Enable Futures
- Hyperliquid: Visit Hyperliquid App → API Settings
Recommended: Yes, use a subaccount dedicated to NOFX for better risk isolation. However, note that some subaccounts have restrictions (e.g., 5x max leverage on Binance).
Common reasons:
- AI decided to "wait" due to market conditions
- Insufficient balance or margin
- Position limits reached (default: max 3 positions)
- See detailed diagnostics in Troubleshooting Guide
Configurable! Default is every 3-5 minutes. Too frequent = overtrading, too slow = missed opportunities.
Yes! You can:
- Adjust leverage settings
- Modify coin selection pool
- Change decision intervals
- Customize system prompts (advanced)
Default: 3 positions. This is a soft limit defined in the AI prompt, not hard-coded. See decision/engine.go:266.
Error: Order's position side does not match user's setting
Solution: Switch to Hedge Mode (双向持仓)
- Login to Binance Futures
- Click ⚙️ Preferences (top right)
- Select Position Mode → Hedge Mode
⚠️ Close all positions first
Why: NOFX uses PositionSide(LONG/SHORT) which requires Hedge Mode.
See Issue #202 and Troubleshooting Guide.
Solution:
# Check what's using port 8080
lsof -i :8080
# Change port in .env
NOFX_BACKEND_PORT=8081Quick Check:
# Is backend running?
curl http://localhost:8080/api/health
# Should return: {"status":"ok"}If not, check Troubleshooting Guide.
Solution:
# Stop all NOFX processes
docker compose down
# OR
pkill nofx
# Restart
docker compose up -d- DeepSeek (recommended for cost/performance)
- Qwen (Alibaba Cloud Tongyi Qianwen)
- Custom OpenAI-compatible APIs (can be used for OpenAI, Claude via proxy, or other providers)
Depends on your model and decision frequency:
- DeepSeek: ~$0.10-0.50 per day (1 trader, 5min intervals)
- Qwen: ~$0.20-0.80 per day
- Custom API (e.g., OpenAI GPT-4): ~$2-5 per day
Estimates based on typical usage. Actual costs vary by provider and usage.
Yes! Each trader can use a different AI model. You can even A/B test different models.
Yes, to some extent. NOFX provides historical performance feedback in each decision prompt, allowing the AI to adjust its strategy.
All data is stored locally in PostgreSQL (Docker volume postgres_data) plus:
decision_logs/- AI decision records
API keys are stored in local databases. Never share your databases or .env files. We recommend using API keys with IP whitelist restrictions.
Yes! Use pg_dump or psql to export data:
docker compose exec postgres \
psql -U nofx -d nofx -c "SELECT * FROM trades;"See the comprehensive Troubleshooting Guide for:
- Step-by-step diagnostics
- Log collection methods
- Common error solutions
- Emergency reset procedures
- Check Troubleshooting Guide first
- Search existing issues
- If not found, use our Bug Report Template
Yes! We welcome contributions:
- Bug fixes and features
- Documentation improvements
- Translations
- See Contributing Guide
Open a Feature Request with your idea!
Last Updated: 2025-11-02