Skip to main content

Python SDK

Use the Python SDK when you want Quantago to call strategy logic that stays in Python.

This is the preferred path when your indicators, research code, or feature engineering already depend on Python libraries.

What the SDK Does​

  • maps the Quantago strategy payload into typed Python dataclasses
  • lets you implement on_candle(...) as normal Python code
  • serves POST /signal and GET /health

Install​

cd sdks/python
python3 -m pip install -e .

Minimal Strategy​

from quantago import Signal, Strategy, StrategyInput


class MovingAverageStrategy(Strategy):
    def on_candle(self, input: StrategyInput) -> Signal:
        closes = [candle.close for candle in input.history]
        if len(closes) < 20:
            return Signal.hold("Need at least 20 candles")

        sma = sum(closes[-20:]) / 20
        if input.candle.close > sma:
            return Signal.buy(size=0.25, reason="Close above 20-period SMA")
        if input.candle.close < sma:
            return Signal.sell(size=0.25, reason="Close below 20-period SMA")
        return Signal.hold("No crossover")

Serve the Strategy​

quantago serve my_strategy:MovingAverageStrategy --host 127.0.0.1 --port 8000

That server exposes:

  • POST /signal
  • GET /health

Connect It to Quantago​

Register it as a remote strategy or pass the runtime metadata in a backtest request:

{
  "type": "remote",
  "language": "python",
  "endpoint": "http://127.0.0.1:8000/signal",
  "timeoutMs": 5000,
  "lookbackCandles": 60
}

Operational Advice​

  • keep the response pure and deterministic for the same input payload
  • size your lookbackCandles to the largest window your indicators need
  • return HOLD instead of raising errors for normal warm-up conditions
  • use the Quantago platform for persistence and orchestration, not the strategy server

When Python Is the Right Choice​

Choose Python when:

  • you already have strategy code in pandas, numpy, or similar tooling
  • you want to iterate on strategies without rebuilding the platform
  • you are comfortable operating a small HTTP service for strategy execution