Manque — an open-source algo-trading framework
A Go framework that closes the gap between a Python strategy prototype and a production C++ version.
The task
Quantitative strategies are usually prototyped in Python, then rewritten in C++ or Rust for production to get the latency down — so the same logic ends up living in two versions that drift apart in bugs and behavior. We needed a tool that doesn't force a choice between research speed and execution speed.
The solution
We built a Go framework with a single path from market data to order: indicators, strategy and risk-gateway behave identically whether running on historical bars in a backtest or on a live exchange WebSocket feed. We added connectors for Binance, OKX, Bybit, Huobi and Coinbase, risk control with a portfolio-wide exposure limit across multiple instruments at once, and an observability stack on Prometheus and Grafana built on top of ready-made metrics.
The result
- The same strategy code runs in backtest, live mode, and across a multi-instrument portfolio without rewrites
- The zero-allocation hot path is verified by CI tests, not just documented in comments
- The risk-gateway caps aggregate exposure across all symbols, not just per instrument
- A public repository with GitHub Actions CI: build, race-detector tests and Docker image smoke tests on every push