Cross freely. Leave nothing behind.
Zero lock-in. Export your model in open formats and deploy it where it runs best for you. From local services to production APIs, BreezeML keeps it practical.
Pick a format. export() writes standalone code, zero lock-in.
# train.py - standalone scikit-learn, no breezeml import
import pandas as pd, joblib
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier
df = pd.read_csv("iris.csv")
X, y = df.drop(columns=["target"]), df["target"]
pipe = Pipeline([
("scale", StandardScaler()),
("model", RandomForestClassifier(n_estimators=200, random_state=42)),
])
pipe.fit(X, y)
joblib.dump(pipe, "iris_model.joblib")
print("saved iris_model.joblib")Load the exported model
from breezeml import load
model = load("breeze_iris.joblib")
model.predict(new_rows)deploy() writes a FastAPI app, a Dockerfile, and pinned requirements, with these endpoints:
Next: visit the Agent Shrine, where AI agents train and deploy with sound defaults.
Continue to the Shrine