Introduction

A pipeline is the named workflow you configure to ingest documents. Each pipeline holds its configuration (a visual graph of nodes and edges), the documents you submit to it, and a history of processing runs.

​

What is a pipeline?

Think of a pipeline as a named workflow for processing a specific type of document. For example, you might create separate pipelines for “Invoice Processing”, “Receipt Scanning”, or “Contract Analysis”. You configure the pipeline once in the pipeline editor, then send documents to it through any of its triggers.

​

Properties

​

Structure

The pipeline configuration is a directed acyclic graph (DAG) made up of:

• Nodes: individual processing steps (triggers, actions, outputs)
• Edges: connections between nodes that define the flow of data

Data flows from trigger nodes through action nodes and out through output nodes. The graph must be acyclic. You cannot create loops.

​

Validation rules

Before you can activate a pipeline, its configuration must pass validation:

• At least one trigger node is required
• All nodes must be connected: no orphaned nodes
• The graph must be acyclic: no circular dependencies
• Each node must have valid configuration (required fields filled in)
• Edges must connect compatible node types (for example, a trigger cannot connect directly to another trigger)

​

Activation

A pipeline must be activated before it can process documents. Activation requires a valid configuration: at least one trigger, all nodes connected, and validation passing. You can deactivate a pipeline at any time to stop processing without deleting it.

​

Versioning

When you save changes to a pipeline, a new version is created. In-flight runs continue using the version that was current when the run started. New runs use the latest saved version.

​

Relationships

• A pipeline has many documents (uploaded files)
• A pipeline has many runs (one per document submitted)
• Documents submitted to a pipeline trigger runs based on its configuration

​

Visual editor

You build and edit pipelines using the visual pipeline editor. The editor provides a drag-and-drop canvas where you add nodes, connect them with edges, and configure each node’s properties.

​

Node types

Pipelines support three categories of nodes:

• Triggers: define how documents enter the pipeline (upload, email, HTTP)
• Actions: process documents (extract, parse, merge, review, route)
• Outputs: deliver results (callback, email, forward)

​

Managing pipelines

You manage pipelines from the Pipelines page. You can create, edit, activate, deactivate, and delete pipelines.