Building a Pipeline

Chain multiple operations together into a visual, editable transformation workflow.

What Is a Pipeline?

A pipeline is a sequence of data transformations. Each operation you apply becomes a card at the bottom of the screen. Cards execute in order, and each one transforms the output of the previous step.

Under the hood, each step is a SQL query. They compose as a chain of Common Table Expressions (CTEs).

Building Your First Pipeline

  1. Upload a file and inspect it in the grid
  2. Click + to open the operation picker
  3. Choose an operation (e.g., Filter)
  4. Configure it in the function panel
  5. Preview the SQL to verify
  6. Click Apply - a card appears in the pipeline
  7. Repeat for each transformation step

Recommended Order

A typical pipeline follows this pattern:

  1. Filter - reduce to the rows you care about
  2. Transform - clean, convert, and compute new columns
  3. Aggregate - summarize or pivot the data
  4. Select - pick and reorder the final columns

Editing Pipeline Steps

  • Edit: Click a pipeline card to re-open its configuration panel with the current settings
  • Delete: Click the X on a card to remove that step

When you edit or delete an earlier step, all subsequent steps are automatically re-executed.

Using Natural Language

Type a description in the pipeline input field (e.g., "remove rows where revenue is negative") and the AI generates a SQL step. You can edit the generated SQL before applying. Requires an API key in Settings.

Tips

  • Start simple - build one step at a time and verify each result
  • Use the SQL preview to catch errors before applying
  • Pipeline steps persist across browser sessions
  • Each view has its own independent pipeline