Pipeline node reference

Builder groups nodes into Capture & Work, Transform & Safety, Quality & Gate, and Exit & Flow. A node is available only when it is shown as available in the current node menu.

Capture & Work

Data Source

Brings existing data into the Pipeline from CSV, File batch, Webhook, or another source available on the page. Select and save the Source type before connecting it to a human-work or processing node.

Supply

Creates Contributor Tasks. Configure Task type, Form template, Reward, Concurrent tasks per user, Limit per contributor, and Items to collect.

Transform & Safety

Agent Task

Uses a wired Agent to process or inspect data. Select an Agent, then configure Instructions, Checks, and the minimum passing score.

Use Transform, De-identify, or Policy Gate only when the current node menu makes it available, and configure the fields shown by the page.

Quality & Gate

Entry Gate

Applies Agreements, courses, Contributor profile requirements, or invite codes before work starts. A Pipeline can have at most one Entry Gate, and it cannot have an incoming connection.

Uniqueness

Checks for duplicate results using selected fields and a matching method. Multiple Rules are evaluated with OR; fields inside one Rule are evaluated with AND.

Consensus

Compares or aggregates results from multiple Contributors. Select the Supply being aggregated, the contributor count, required passing count, and fields that must agree.

Exit & Flow

Accept

Performs final acceptance. Configure an available Auto, Manual, Scripted, or Agent method. The minimum launch path can end at Accept; connect it to ETL or Sink when the Campaign must deliver a Dataset.

ETL

Maps accepted results to a stable delivery structure without changing the original Submission. It can rename fields and apply defaults, fallbacks, value mappings, composed fields, or fixed fields.

Sink

Writes accepted records to the Campaign Dataset. It normally follows Accept or ETL.

Routes

Configures a supported branch, fallback, bounded loop, or sample. Standard node connections must pass Builder validation; do not draw an unsupported loop manually.

Common paths

Minimum launchable human workflow:

Entry Gate → Supply → Accept

Direct collection with Dataset delivery:

Entry Gate → Supply → Accept → Sink

Quality checks and field transformation:

Entry Gate → Supply → Uniqueness / Consensus → Accept → ETL → Sink

Label and validate an existing CSV:

Entry Gate → Data Source → Supply (label) → Supply (validate) → Accept → Sink

Save each node after configuring it, then save the complete Pipeline graph.