Data Sink¶

You want to save results to a file - scoring a full dataset and writing the output to parquet or CSV for downstream analysis.

Spreadsheet equivalent

Like "Save As CSV" in Excel, but built into your pipeline so it runs automatically.

When to use

Use this for batch scoring - processing a full dataset and saving the results. For live API responses, use Output instead.

This node accepts a single input.

Config	Description
`path`	Required. Output file path (e.g. `outputs/scored_policies`)
`format`	Required. `"parquet"` or `"csv"`

If you provide a filename without a directory, it's written to outputs/. The format extension is added automatically if missing.

Example:

{
  "path": "outputs/scored_policies",
  "format": "parquet"
}

This writes the full scored dataset to outputs/scored_policies.parquet.

All columns are written

All columns from the input are written to the file. To control which columns are saved, add a Polars node upstream with selected_columns.

Overwrites existing files

If the file already exists, it is overwritten.

Multiple sinks

You can have multiple Data Sink nodes in a pipeline - for example, to write both a parquet file and a CSV, or to save results at different stages.

See also: