andlake 0.1.1

Andlake platform SDK — pre-configured connections to Trino, Nessie, MLflow, S3, and Iceberg.

andlake-sdk

Pre-configured Python SDK for the Andlake data platform. Provides zero-config access to Trino, Nessie, MLflow, S3, and Iceberg from JupyterHub notebooks.

Quick Start

from andlake import get_trino_connection, configure_mlflow
import pandas as pd

# Connect to Trino via the Andlake gateway
conn = get_trino_connection()
df = pd.read_sql("SELECT * FROM lake.silver.transactions LIMIT 1000", conn)

# Set up MLflow experiment tracking
configure_mlflow(experiment_name="fraud-detection")

Available Functions

Function	Description
get_trino_connection()	Trino DBAPI connection via the gateway
get_trino_engine()	SQLAlchemy engine for pd.read_sql()
get_nessie_client()	Nessie catalog client for branch management
configure_mlflow()	Set MLflow tracking URI and experiment
get_mlflow_client()	Pre-configured MlflowClient
get_s3_client()	boto3 S3 client (uses IRSA)
get_s3_resource()	boto3 S3 resource (uses IRSA)
get_iceberg_catalog()	PyIceberg REST catalog via Nessie

Environment Variables

Static service URLs are set by JupyterHub extraEnv. Per-user values are injected by the pre_spawn_hook from Keycloak auth_state.

Variable	Default	Source
ANDLAKE_GATEWAY_URL	http://notebook-service:8082	extraEnv
TRINO_HOST	notebook-service	extraEnv
TRINO_PORT	8082	extraEnv
NESSIE_URI	http://nessie:19120/api/v2	extraEnv
MLFLOW_TRACKING_URI	http://mlflow:5000	extraEnv
ANDLAKE_DEFAULT_CATALOG	lake	extraEnv
ANDLAKE_S3_BUCKET	andlake-app	extraEnv
ANDLAKE_TENANT_ID	(required)	pre_spawn_hook
ANDLAKE_ACCESS_TOKEN	(required)	pre_spawn_hook

Development

pip install -e ".[dev]"
pytest