SDK Reference
Complete API surface for the inferpathio Python package. For quickstart usage, see the Quickstart guide.
ifp.init()
Initialize the SDK for a process. Must be called before any other SDK method.
ifp.init(
api_key: str,
# Optional: override API base URL (for on-prem deployments)
base_url: str = "https://api.inferpathio.com/v1"
)ifp.ModelTracker
The primary class for continuous drift monitoring and governance event creation.
tracker = ifp.ModelTracker(
model_id: str,
policy: str, # policy name registered in dashboard
version: str = None # optional version tag
)
# Log a prediction batch for drift monitoring
tracker.log_prediction_batch(
features: pd.DataFrame | np.ndarray,
labels: pd.Series | np.ndarray,
threshold: float = 0.1,
metric: str = "psi" # "psi" | "kl" | "wasserstein"
)ifp.governance_run()
Context manager for attaching governance metadata to a model training or promotion run.
with ifp.governance_run(
model_id: str,
metadata: dict = None # custom key-value tags
) as run:
run.log_model(model: Any, flavor: str)
run.set_policy(policy_name: str)
run.tag(key: str, value: str)
run.log_metric(name: str, value: float)
run.promote(environment: str) # triggers approval workflowSupported model flavors
'sklearn'— scikit-learn Pipeline or estimator'xgboost'— XGBoost Booster'lightgbm'— LightGBM Booster'pytorch'— PyTorch nn.Module'tensorflow'— Keras/TF SavedModel'mlflow'— MLflow logged model URI'custom'— any callable with predict()