Quickstart Guide

This guide walks you through creating a project, building a simple pipeline, seeding tasks, and completing work as a contributor.

Prerequisites

• A Pipelines account with Project Admin (owner) or Org Admin permissions

Step 1: Create a project

1. Navigate to Projects in the sidebar.
2. Click Create Project.
3. Enter a name for your project.
4. Click Create. You will be redirected to the new project's dashboard.

Step 2: Build a pipeline

1. Inside your project, go to Pipelines in the sidebar.
2. Click Create Pipeline (or choose Create from Template to start with a pre-built pipeline).
3. The Pipeline Builder opens — a visual canvas where you design your pipeline.

Add a subtask node

1. The canvas starts with a Start node and an End node.
2. Click the + button between them or drag a new Subtask node from the toolbar.
3. Click the subtask node to open its Form Builder in the side panel.
4. Add fields to the form. Available types include Short Text, Long Text, Code, Select, Rating, Pairwise, Ranking, File Upload, and more. See Form Builder & Field Types for the full list.
5. Configure each field. Every field has a rich set of options:
   • Input mode — choose whether the field is filled by a contributor, pre-populated with seeded data, set to a static value, or auto-generated by an LLM.
   • LLM configuration — for AI-generated fields, select a model, write a prompt (or pull one from the Prompt Library), configure tool-calling, and control which upstream fields provide context.
   • Validation & display — set character/word limits, min/max values, required vs. optional, grid width, conditional visibility based on other field values, and more.
   • Evaluation — attach evaluation criteria directly to fields for inline quality scoring.

Add a review node (optional)

1. Drag a Review node onto the canvas between the subtask and the end node.
2. Connect the subtask node's output to the review node's input.
3. Connect the review node's output to the end node.
4. Click the review node to configure which subtask fields should be reviewed.

Connect the nodes

Drag from any node's output handle to another node's input handle to create edges. The start node should connect to your first subtask, and the last node should connect to end.

Step 3: Publish the pipeline

1. Once your pipeline graph is complete, click Publish in the top-right corner of the builder.
2. The pipeline status changes from Draft to Active.
3. The pipeline is now ready to accept tasks.

Step 4: Seed tasks

Tasks are the units of work that flow through your pipeline. You can create them in several ways:

From CSV

1. Go to your pipeline's Data Explorer (click the pipeline name in the sidebar).
2. Click Create Tasks and select Upload CSV.
3. Map CSV columns to your subtask form fields.
4. Upload the file. Tasks are created in the background.

From the API

curl -X POST https://api.pipelines.tech/api/v1/workflows/{workflow_id}/tasks \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "tasks": [
      {
        "fields": {
          "input_text": "Translate this sentence to French: Hello, world!"
        }
      }
    ]
  }'

Infinite seeding

Enable Infinite Seeding in pipeline settings to generate empty tasks on demand whenever a contributor needs work.

Step 5: Complete work as a contributor

1. Navigate to Tasks to see your work queue.
2. Click Claim on an available task.
3. Fill out the form fields and click Submit.
4. If a review node follows, the task moves to a reviewer's queue.

Step 6: View results

1. Click the pipeline name in the sidebar to go back to the Data Explorer.
2. Browse completed tasks in the table view — the Metadata and Content tabs show task data, with additional Evaluation Analytics and LLM Analytics tabs available when those features are enabled.
3. Export results as CSV, JSON, or ZIP.

Next steps