MetaTrader5 MCP Server

Sktime Adapter ============== The forecasting stack includes a generic adapter for sktime forecasters, exposed as the `sktime` method. Requirements - `sktime` (installed via `pip install -r requirements.txt`) Usage - Method name: `sktime` - Params: - `estimator` (str): fully qualified class path, e.g., `sktime.forecasting.theta.ThetaForecaster` - `estimator_params` (object): constructor kwargs - `seasonality` (int, optional): injected as `sp` if supported by the estimator Examples ```bash # ThetaForecaster via sktime python cli.py forecast_generate EURUSD --timeframe H1 --method sktime --horizon 16 \ --params '{"estimator":"sktime.forecasting.theta.ThetaForecaster"}' # NaiveForecaster (last) with seasonal period 24 python cli.py forecast_generate EURUSD --timeframe H1 --method sktime --horizon 24 \ --params '{"estimator":"sktime.forecasting.naive.NaiveForecaster","estimator_params":{"strategy":"last"},"seasonality":24}' # AutoETS with additive errors python cli.py forecast_generate EURUSD --timeframe H1 --method sktime --horizon 24 \ --params '{"estimator":"sktime.forecasting.ets.AutoETS","estimator_params":{"error":"add"}}' ``` Notes - If `estimator` is omitted, the adapter defaults to `NaiveForecaster(strategy="last")`. - When `seasonality` is provided and the estimator accepts `sp`, the adapter injects it automatically. - If you pass exogenous features via the existing `features` mechanism, they are forwarded to `fit(X=...)` and `predict(X=...)` when available.