Quickstart
This guide gets you from zero to a working Inferpathio experiment in under 15 minutes. By the end you'll have a tracked training run, a registered model version, and drift monitoring configured.
Prerequisites
- Python 3.8 or later
- An Inferpathio account — sign up free
- Your API key from the account settings
Step 1: Install the SDK
pip install inferpathio
Step 2: Authenticate
ifp configure --api-key ifp_sk_your_key_here
This writes your credentials to ~/.ifp/credentials. Alternatively, set the environment variable IFP_API_KEY.
Step 3: Track your first experiment
import inferpathio as ifp
import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
X, y = make_classification(n_samples=5000)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
with ifp.track("my-first-model") as run:
clf = RandomForestClassifier(n_estimators=100, max_depth=8)
clf.fit(X_train, y_train)
acc = clf.score(X_test, y_test)
run.log_metric("accuracy", acc)
run.log_param("n_estimators", 100)
run.log_param("max_depth", 8)
run.register(stage="staging")
Step 4: Set up drift monitoring
from inferpathio import monitor
m = monitor("my-first-model")
m.set_baseline(X_train) # training distribution as reference
m.drift_threshold = 0.10 # PSI threshold
m.on_drift(action="alert", channel="email")
m.watch(X_live) # pass production batches here
What's next?
- Full Python SDK reference — all methods and parameters
- Drift monitoring configuration — KS test, Wasserstein distance, custom metrics
- Automated retraining — trigger policies with YAML
- Integrations — connect Airflow, Kubeflow, GitHub Actions