​

Task Agents are Workflow components in which a custom operation on all tasks in the Agent stage can be triggered. This allows you to set up pre-labeling, like using foundation models such as GPT-4o, automated quality assurance, or any other custom action you need for your workflow. Here are some common use-cases:
  • Pre-labeling of your data, for example, with your own model or off-the-shelf models.
  • Custom routing of data in the Project Workflow based on, for example, metadata, annotation time, or label counts.
  • Dynamic prioritization of your tasks. This lets you “rearrange” tasks before sending them to review for example.
  • Custom “label assertions” that validate, for example, number of labels or other constraints, before sending them for review.
  • Custom consensus computations by reading in labels from other Projects.
Task Agents must be triggered when a task reaches the associated agent state in the Project Workflow.
Learning to build Task Agents is best done using examples.