DuckLake

DuckLake is an open lakehouse format with a SQL catalog database (PostgreSQL, DuckDB, SQLite) and Parquet data files on any storage backend (local, S3, R2, GCS, Azure).

Every mutation creates an immutable snapshot, enabling time travel, change feeds, and conflict resolution.

Installation & Connection

INSTALL ducklake;
LOAD ducklake;

-- Local DuckDB catalog + local files
ATTACH 'ducklake:metadata.ducklake' AS my_lake;

-- PostgreSQL catalog + S3 storage
ATTACH 'ducklake:postgres:dbname=ducklake host=myhost' AS my_lake
    (DATA_PATH 's3://my-bucket/data/');