DuckLake Python SDK

The ducklake Python package provides a Pythonic interface for interacting with DuckLake — an open table format for lakehouses, backed by a SQL catalog database and Parquet files in object storage.

It is built on top of a Rust implementation of the DuckLake specification and exposes first-class integrations with DuckDB, Polars, and Apache Arrow.

Warning

This is not an official SDK released by the DuckDB Foundation. The ducklake name on PyPI was reserved by the DuckDB Foundation, hence the package is distributed as ducklake-sdk on PyPI but importable as ducklake.

Installation

Install the package from PyPI using your favorite package manager:

pip install ducklake-sdk
pixi add ducklake-sdk

To enable the optional integrations, install one of the extras:

pip install "ducklake-sdk[arrow]"     # Arrow + DuckDB based reads/writes
pip install "ducklake-sdk[polars]"    # Polars based reads/writes
pip install "ducklake-sdk[sqlalchemy]" # SQLAlchemy URLs as catalog connection strings

Quick Example

import ducklake as dl
import polars as pl

# Create a new DuckLake with local storage
ducklake = dl.create("sqlite:///metadata.sqlite", data_path="data_files/")

# Create a new table
table = ducklake.create_table(
    "test",
    schema={"i": dl.Int64(), "s": dl.Varchar()},
)

# Insert data into the table
lf = pl.LazyFrame({"i": [1, 2, 3], "s": ["hello", "ducklake", "sdk"]})
table.sink_polars(lf)

# Read data from the table
lf = table.scan_polars()