Platform Guides
Workflows
Orchestrate labeling and LLM-driven workflows and track their individual tasks.
Overview
A workflow is a reusable pipeline that turns input data into labeled output. The platform supports two kinds: labeling workflows, which route work to human annotators, and LLM workflows, which run prompts against a model. A workflow run produces a set of workflow tasks — the individual units of work.
Labeling workflows
| GET | /api/v1/labeling-workflows | List labeling workflows. |
| POST | /api/v1/labeling-workflows | Create a labeling workflow. |
| POST | /api/v1/labeling-workflows/{id}/run | Start a run of a labeling workflow. |
LLM workflows
LLM workflows define a prompt template and model configuration applied across a dataset.
| GET | /api/v1/llm-workflows | List LLM workflows. |
| POST | /api/v1/llm-workflows | Create an LLM workflow. |
Workflow tasks
Tasks are emitted as a workflow run progresses. Poll them to monitor completion or fetch results.
| GET | /api/v1/workflow-tasks | List tasks, filterable by workflow run. |
| GET | /api/v1/workflow-tasks/{id} | Retrieve a single workflow task. |
Continue with Labeling & Tasks for task dispatch and auto-labeling, or open the Label Backend API reference.