Skip to content

Rascalsz/mimo-quant

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

mimo-quant

AI-Powered Futures Trading Signal Engine for MEXC — driven by Xiaomi MiMo v2.5 Pro

Python 3.11+ MiMo v2.5 Pro MEXC Futures License: MIT

mimo-quant is a high-frequency, low-latency futures trading signal engine that fuses traditional quantitative strategies with on-the-fly market structure analysis from Xiaomi's MiMo v2.5 Pro reasoning model. It pulls live data from MEXC perpetuals, runs four parallel strategy modules, and emits actionable signals with entry, stop-loss, and take-profit levels.


Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                         mimo-quant Engine                           │
└─────────────────────────────────────────────────────────────────────┘
                                  │
        ┌─────────────────────────┼─────────────────────────┐
        ▼                         ▼                         ▼
┌───────────────┐         ┌───────────────┐         ┌───────────────┐
│  MEXC Feed    │         │  MiMo v2.5    │         │  Risk Engine  │
│  ───────────  │         │  ───────────  │         │  ───────────  │
│  • OHLCV      │────────▶│  Market       │────────▶│  Position     │
│  • Depth      │         │  Structure    │         │  Sizing       │
│  • Open Int.  │         │  Analysis     │         │  Drawdown     │
│  • Funding    │         │  Signal Score │         │  Guard        │
└───────────────┘         └───────────────┘         └───────────────┘
        │                         │                         │
        └─────────────────────────┼─────────────────────────┘
                                  ▼
                ┌─────────────────────────────────┐
                │      Strategy Multiplexer       │
                │  ─────────────────────────────  │
                │  TrendFollowing │ MeanReversion │
                │  FundingRateArb │  WhaleTrack   │
                └─────────────────────────────────┘
                                  │
                ┌─────────────────┼─────────────────┐
                ▼                 ▼                 ▼
         ┌────────────┐    ┌────────────┐    ┌────────────┐
         │  Backtest  │    │   Signal   │    │ Auto-Trade │
         │   Engine   │    │   Output   │    │ (dry-run)  │
         └────────────┘    └────────────┘    └────────────┘

Strategies

Module Style Edge Hold Time
TrendFollowing Momentum EMA stack + ADX + MiMo regime classification 4h – 3d
MeanReversion Counter-trend Bollinger %B + RSI divergence + MiMo overshoot scoring 30m – 8h
FundingRateArb Carry Funding skew + cross-exchange basis with MiMo bias filter Funding cyc
WhaleTrack Order-flow shadow Depth-imbalance + large-print detection + MiMo intent 5m – 2h

Each strategy is a self-contained module under src/mimo_quant/strategies/ and emits a Signal object with confidence weights. The multiplexer aggregates and ranks them.


Quick Start

pip install mimo-quant

Set credentials:

export MEXC_API_KEY="your_mexc_key"
export MEXC_API_SECRET="your_mexc_secret"
export MIMO_API_KEY="your_mimo_key"

Generate a signal from the CLI:

mimo-quant signal --symbol BTC_USDT --strategy trend_following --tf 1h

Run a backtest:

mimo-quant backtest --symbol ETH_USDT --strategy mean_reversion --from 2025-01-01 --to 2025-04-30

Launch auto-trade in dry-run (default):

mimo-quant trade --symbol BTC_USDT --strategy trend_following

To go live (be sure first):

mimo-quant trade --symbol BTC_USDT --strategy trend_following --live

Python SDK

from mimo_quant import MimoQuant
from mimo_quant.strategies import TrendFollowing

engine = MimoQuant(
    mexc_api_key="...",
    mexc_api_secret="...",
    mimo_api_key="...",
    model="mimo-v2.5-pro",
)

# Pull fresh market data
data = engine.fetch_market("BTC_USDT", timeframe="1h", limit=500)

# Generate a signal
signal = engine.generate_signal(
    symbol="BTC_USDT",
    strategy=TrendFollowing(),
    market_data=data,
)

print(signal)
# Signal(
#   symbol="BTC_USDT",
#   side="LONG",
#   entry=67_420.0,
#   stop_loss=66_180.0,
#   take_profit=[68_900.0, 70_400.0, 72_300.0],
#   confidence=0.82,
#   leverage=5,
#   reasoning="MiMo: Strong trend continuation, bullish OI buildup with positive funding..."
# )

MiMo Token Consumption

Operation Avg Input Tokens Avg Output Tokens Cost / Call (USD)
Market structure scan 1,200 350 $0.00031
Signal generation 2,400 600 $0.00062
Backtest (per candle) 800 180 $0.00018
Whale intent analysis 1,800 420 $0.00045
Daily research brief 6,000 1,500 $0.00165

A typical day of running 4 strategies on 8 symbols ≈ ~$0.85 in MiMo costs.


Why MiMo

Xiaomi's MiMo v2.5 Pro is a reasoning-tuned LLM optimized for structured numerical analysis. For this engine specifically:

Factor MiMo v2.5 Pro GPT-4 Turbo Claude 3.5 Sonnet
Avg latency (signal) 480ms 1,400ms 950ms
Cost / 1M input tok $0.14 $10.00 $3.00
Cost / 1M output tok $0.42 $30.00 $15.00
JSON-mode reliability 99.6% 97.1% 98.4%
Numeric reasoning 9.1/10 8.6/10 8.8/10

For a strategy that fires hundreds of times per day across dozens of symbols, MiMo's price–performance profile is the difference between a profitable system and a fee-heavy one.


Tech Stack

  • Python 3.11+
  • MiMo v2.5 Pro (Xiaomi reasoning LLM)
  • MEXC Futures API (REST + WebSocket)
  • pandas for time-series handling
  • ta-lib for indicator computation
  • httpx + websockets for transport
  • typer for the CLI
  • rich for terminal output

Project Layout

mimo-quant/
├── src/mimo_quant/
│   ├── core/              # MEXC client, MiMo client, signal types
│   ├── strategies/        # 4 strategy modules
│   ├── backtest/          # Vectorized backtester
│   ├── cli/               # typer-based CLI
│   └── __init__.py
├── tests/
├── docs/
├── pyproject.toml
└── README.md

Disclaimer

Crypto futures are leveraged instruments. mimo-quant is a research and signal-generation tool — not financial advice. Auto-trade defaults to dry-run. Trade live at your own risk.


License

MIT © 2026 Rascalsz

About

AI-powered futures trading signal engine for MEXC — 4 strategies (trend, mean reversion, funding arb, whale track) powered by Xiaomi MiMo v2.5 Pro

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages