A flow is a chain of stages: sources on the left, transformations in the middle, writers at the end.
Create a flow
Open the library's overflow menu (⋮, titled "ETL Builder actions") and choose "Create flow" — the same menu offers "Open flow". Give it a "Flow name" and an optional "Description", then click "Create flow".
The empty canvas tells you where to start: "Drag a stage from the library on the left — start with a source like Dataset or Read Data."
Add stages
The stage library runs down the left side, grouped into categories with a "Search stages..." box at the top. Drag a stage onto the canvas, or double-click it.
Connect them
Drag from one stage's output port to the next stage's input port, or click one port and then the other. To rewire, grab the end of an existing connection and drop it somewhere else.
Stages only accept connections that make sense. Readers have no input, writers have no output, and anything in "Joins" needs two inputs before it will run.
Organize the canvas
- Sections. Click "Add section" to draw a frame around related stages. Its menu offers "Rename section", "Fit section to contents", and "Delete section" — handy for labelling the ingest, transform, and output parts of a bigger flow.
- Auto arrange. Tidies the whole graph in one click, from the canvas top bar.
- Node style. In the ⋮ menu, "Node style" switches between "Detailed", "Compact", and "Classic". Compact fits more graph on screen.
Save your work
"Save" and "Close flow" sit in the top bar, and an "Unsaved changes" chip appears whenever the canvas is ahead of what's saved. Save before you run — and before "Publish to Server", which works from the saved flow.