Use your own data

Load a file, check its columns, transform it and download a CSV. You need a project you can edit and run (Editor or Project Admin access). If you have not signed in yet, start with installation; for an introduction to SQL nodes, follow the quickstart.

Choose a source

Project files are uploaded to the Duckpad server. A path on your laptop is not a path on a remote server. Upload the file, or ask an instance administrator to configure a connection to storage the server can reach. See connections for database setup, formats and permissions.

1. Upload and inspect a CSV

On Projects, choose New project → Empty project, give it a name and choose Create. Keep dev selected for this example. Save this text as orders.csv on your computer, or use your own CSV:

order_id,region,amount
001,Europe,120
002,Europe,30
003,North America,80

Drop the file onto the Flow canvas. Duckpad uploads it and creates a reader node named after the file. You can also click Load CSV, leave Location set to Project files, and choose Upload file in the inspector.

The inspector shows the first 20 rows before a run. Check the delimiter and header settings. For this example, set the order_id column’s data type to VARCHAR to keep leading zeros. Auto uses DuckDB’s detected type. Changing the delimiter or header clears type overrides, so review them afterward.

Choose Run all, then select the node. Results → Data should show three orders, with IDs 001, 002 and 003. A source preview in the inspector does not publish a result; a run does.

Uploads are limited to 1 GiB per file. For larger files, ask an administrator for a Server folder connection. For XLSX, select the sheet and check the workbook limits.

2. Transform the rows

Click Select columns in the palette. Connect the reader’s output to its source input, either by dragging between ports or using the inspector’s input selector. The upstream run makes column choices available.

Choose Use all columns to add order_id, region and amount. In the amount row, change Output name to revenue. Choose Run with ancestors on this node. Its Data tab should show the same three orders with the new column name. For other operations, use Clean values or write SQL.

Filters and samples in Results → Data only change the displayed rows. Use a Filter node to change what downstream nodes and exports receive. You can turn a sample filter into a node with Add as a Filter node.

3. Export and download

1. Click Export CSV in the palette.
2. Connect the transformation’s output to the export’s source input.
3. Keep Location → Project files, set destination to revenue.csv, and choose the delimiter and header settings you want.
4. Choose Run with ancestors on the export. This reruns its upstream nodes and writes the full input table.
5. Open the export node’s Log and choose Download exported data.

The download contains all three rows, including the revenue column. An export has no output table, so its results panel opens on Log. A repeated CSV export to the same destination replaces that file; a database Export table appends rows instead. Check the selected environment before writing to shared storage.

The results guide explains which exports can be downloaded and when another export is required. For a custom query, see SQL exports.

Keep and share the project

Completed edits are saved as Git commits. Use Code → Download project ZIP to share the pipeline definition, without uploaded files, results or connections (project files).

Use Download exported data to share a result file, and project access to let colleagues work on the same server. A project ZIP is not an instance backup.

To refresh a source, upload and select its new file or update the connected resource, then run again. Edits, file changes and environment switches do not run the pipeline automatically. See result lifetime and troubleshooting if the displayed rows are unexpected.