Skip to main content

Overview

This page is a one-page orientation: what a workflow looks like, the configuration object that defines it, and how its parts connect. Each section links out to the deeper references.

What a workflow looks like​

Four regions to recognise on every workflow page:

  • Topbar — workflow-level controls (export, history, run-all, panel toggles).
  • Navigation — the step tree, with the active step highlighted.
  • Input — the active step's auto-generated form and its run controls.
  • Output — the step's viewer or scalar block.

Annotated workflow page with Topbar, Navigation, Input, and Output regions

Each region maps onto a part of the configuration object covered next.

How it's defined​

A workflow is a TypeScript function — the workflow provider — that returns a PipelineConfiguration. The callouts below link to each field's canonical reference:

//tags: model//editor: Compute2:TreeWizardEditor//input: object params//output: object resultexport function MyWorkflow(params: any): PipelineConfiguration { return { id: 'myworkflow', type: 'static', steps: [ { id: 'loadData', nqName: 'mypkg:LoadData', }, { id: 'analyse', nqName: 'mypkg:Analyse', }, { id: 'subflow', type: 'dynamic', stepTypes: [ { id: 'preprocess', nqName: 'mypkg:Preprocess' }, ], }, ], links: [ { id: 'shareLoaded', type: 'data', from: 'out:loadData/Output', to: 'in:analyse/Input', }, ], };}workflow providerpipeline typescript node (nqName)nested dynamic pipelinelink (data, validator, meta, action)LQL selectors for source / target

Configuration — the central artifact​

A workflow is fully described by a single TypeScript object of type PipelineConfiguration — in code the type and its runtime cousins use the Pipeline* prefix; throughout these docs the prose calls it a workflow. The configuration is a tree: workflows contain steps; steps are either script nodes, nested workflows, or references to other workflows. Every node has an id. The type field discriminates workflows (static, dynamic, action, ref); leaves are scripts (type: 'step', or omitted). See Configuration for the full field-by-field reference.

Links live alongside steps inside a workflow and propagate data between IOs. Data flow is forward-only, in depth-first order through the tree. Some link roles target virtual outputs rather than script IOs — validator for per-IO validation, pipelineValidator for workflow-level checks, meta for UI hints, plus the user-triggered action. See Link types for the catalogue and Link Query Language for how from/to paths are written.