> For the complete documentation index, see [llms.txt](https://docs.unitlab.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.unitlab.ai/documentation/workflows/workflows-overview.md).

# Workflows overview

Route project work through human annotation, automated processing, quality assurance, and completion.

A Unitlab workflow determines where a work item is, who can act on it, which actions are available, and where it goes next. It connects source intake, annotation, automated processing, quality decisions, and completion in one visible process.

Use the workspace Workflows area to manage reusable workflows. Open a project's **Workflows** tab to configure its active workflow. Use queues to operate assigned work and the project's **QA** area to inspect quality evidence.

## Start with a clear operating path

```mermaid
flowchart LR
  P[Project] --> A[Annotate]
  A --> R[Review]
  R -->|Approved| C[Complete]
  R -->|Rejected| A
```

This basic process separates creating labels from accepting them. Add stages when the task needs another distinct responsibility, such as independent votes, reference checks, model inference, or specialist review.

## Understand the stage families

| Responsibility         | Available stages                                                    |
| ---------------------- | ------------------------------------------------------------------- |
| Intake                 | Project.                                                            |
| Human work             | Annotate, Review, and Consensus assignments.                        |
| Quality evaluation     | Quality Gate; Consensus compares completed independent submissions. |
| Automation and routing | Model, Logic, Sampling, and Webhook.                                |
| Terminal disposition   | Complete and Archive.                                               |

See [Stages and routes](https://docs.unitlab.ai/documentation/workflows/stages-and-routes) for the catalog, configuration steps, and required outcomes. Specialized names such as Expert Review are configured instances of a supported stage type.

## Add the right QA control

* **Consensus** collects independent submissions and checks whether enough answers agree. It waits for all votes, then follows Pass or sends Fail to a Review stage.
* **Quality Gate** compares with an approved answer key. Configure Pass, Fail, and Not evaluated separately.
* **Review** gives a qualified person responsibility for accepting or returning the work.
* **Quality checks** surface validation problems and unresolved issues in project QA. Inspect and resolve them in the affected item's context.

Read [QA Overview](https://docs.unitlab.ai/documentation/qa/qa-overview) to choose controls and [QA Workflows](https://docs.unitlab.ai/documentation/qa/qa-workflows) to combine them. An agreement score, a reference comparison, and a reviewer decision answer different questions.

## Choose the next guide

| Task                                         | Guide                                                                                                           |
| -------------------------------------------- | --------------------------------------------------------------------------------------------------------------- |
| Build or extend a graph                      | [Stages and routes](https://docs.unitlab.ai/documentation/workflows/stages-and-routes).                         |
| Assign people and understand actions         | [Assignment and stage actions](https://docs.unitlab.ai/documentation/workflows/assignment-and-stage-actions).   |
| Operate corrections and specialist decisions | [Review, rework, and escalation](https://docs.unitlab.ai/documentation/workflows/review-rework-and-escalation). |
| Configure inference                          | [Model stages](https://docs.unitlab.ai/documentation/workflows/model-stages).                                   |
| Modify active operations                     | [Change a live workflow](https://docs.unitlab.ai/documentation/workflows/change-a-live-workflow).               |
| Configure independent voting                 | [Consensus](https://docs.unitlab.ai/documentation/qa/consensus).                                                |
| Configure approved-reference checks          | [Quality Gate](https://docs.unitlab.ai/documentation/qa/quality-gate).                                          |

## Keep these operating boundaries clear

Saving appends annotation history; a stage action changes the item's workflow position. Completing one check does not imply that later checks have finished. Rejection needs a correction destination, while an unavailable comparison needs its own policy. Queue visibility and valid actions follow the current stage, role, assignment, and task state.

A production workflow is ready when representative items can traverse its normal and exception paths, every active task has an owner, and the final acceptance criteria are clear before release.