frozen-ducklake 0.11.0

Frozen DuckLake CLI

fdl — Frozen DuckLake CLI

Manage Frozen DuckLake catalogs — place a DuckLake catalog + Parquet files on object storage and anyone can query with a single DuckDB ATTACH. No database server required.

Quick Start

pip install frozen-ducklake

mkdir my_dataset && cd my_dataset
fdl init my_dataset

Add data, push, and serve:

fdl sql default "CREATE TABLE cities (name VARCHAR, country VARCHAR, pop INTEGER)"
fdl sql default "INSERT INTO cities VALUES ('Tokyo', 'Japan', 14000000), ('Shanghai', 'China', 24900000)"

fdl push default
fdl serve default

Query from DuckDB:

duckdb -c "ATTACH 'ducklake:http://localhost:4001/my_dataset/ducklake.duckdb' AS my_dataset;
           SELECT * FROM my_dataset.main.cities ORDER BY pop DESC;"

Features

• init / push / pull — Git-like catalog management with conflict detection
• sql — Query and modify data directly from the command line
• run — Execute any command (dbt, dlt, Python) with auto-injected storage credentials
• expire — Snapshot retention and storage cleanup, runs automatically after writes
• serve — Built-in HTTP server with CORS and Range request support
• Python API — Call the same operations in-process from Dagster, Airflow, etc.
• Works with S3-compatible storage (AWS S3, Cloudflare R2, GCS, etc) and local directories

Install

pip install frozen-ducklake   # pip
uv tool install frozen-ducklake   # uv (recommended)
pipx install frozen-ducklake   # pipx

Documentation

fdl.flo8s.com

License

MIT