Model Recipe Plugins

The dashboard only lists model recipes registered by trusted Python code on the local machine. It does not execute code pasted into the browser.

Create an installed module such as my_project.pspso_recipes:

from sklearn.ensemble import AdaBoostClassifier

from pspso import FloatRange, IntRange, SearchSpace, register_recipe

register_recipe(
    "ada_boost",
    AdaBoostClassifier,
    tasks=["binary_classification", "multiclass_classification"],
    description="Local AdaBoost recipe for comparison experiments.",
    fixed_params={"binary_classification": {"random_state": 42}},
    search_spaces={
        "binary_classification": SearchSpace({
            "n_estimators": IntRange(25, 150),
            "learning_rate": FloatRange(0.01, 1.0, precision=2),
        })
    },
)

Configure the module before starting the dashboard or CLI:

set PSPSO_RECIPE_PLUGINS=my_project.pspso_recipes
uv run pspso-dashboard

Use a comma-separated module list for several trusted plugin modules. Plugins run on the local backend, so treat them like any other installed Python code.