DuckLake
Connect a DuckLake lakehouse. DuckLake keeps table metadata in a SQL catalog and table data as files in storage; Datyo reads both and mounts the tables into the query engine, where they can be joined with datasets from other connectors.
How DuckLake is addressed
A DuckLake connection has two parts:
- Catalog — a SQL database holding the lakehouse metadata. It can be a DuckDB file, a SQLite file, or a PostgreSQL / MySQL database.
- Data path — where the table data files (Parquet) live: an object-storage location or a local path.
Before you begin
- Access to the catalog: a catalog file (DuckDB/SQLite) reachable by Datyo, or connection details for a PostgreSQL / MySQL catalog database.
- Read credentials for the data path if it is on object storage (see object-storage access credentials).
Connection settings
| Field | Required | Description |
|---|---|---|
| Catalog metadata store | Yes | DuckDB file, SQLite file, PostgreSQL, or MySQL. |
| Catalog file path | For file catalogs | Path/location of the DuckDB or SQLite catalog file. |
| Catalog DSN | For PostgreSQL / MySQL | Connection details for the catalog database. Stored encrypted. |
| Storage location | Yes | Object storage, or a local path where the environment permits it. |
| Data path | Yes | Where the data files live, e.g. s3://bucket/data/. |
For an object-storage data path, also provide the object-storage access credentials.
TIP
On the managed service, use a data path in object storage. A local data path or a local catalog file is available in the desktop app and self-hosted deployments — see Managed vs. self-hosted.
Test and save
Enter the fields and use Test connection. Datyo opens the catalog, then reads the tables it describes from the data path. On success, choose the tables to import.
Capabilities
- Execution model: Federated. DuckLake tables can be joined with datasets from other connectors.
- Access: Read-only.
- Catalog backends: DuckDB file, SQLite file, PostgreSQL, MySQL.
Troubleshooting
- Cannot open the catalog — verify the catalog file location, or the PostgreSQL/MySQL connection details.
- Tables list but data cannot be read — the catalog opened, but the data path credentials are missing or lack access; check the object-storage keys.
- Wrong catalog type — the metadata store selection must match how the lakehouse was created.