> ## Documentation Index
> Fetch the complete documentation index at: https://velurahq.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Quality Gates: approve, score, and route every draft

> Pause workflows at decision points, notify reviewers, and score drafts automatically so your team controls every content decision before production.

Quality gates are the control points inside VeluraHQ workflows that keep humans in the loop. When a workflow reaches a gate, it pauses, notifies the assigned owner, and waits. Nothing moves to the next step — including CMS publishing — until a person reviews and approves. You decide where gates sit, who owns them, and what score threshold triggers them, so the level of human oversight always matches the stakes of the content.

## Types of quality controls

<CardGroup cols={2}>
  <Card title="Human Review Gate" icon="user-check">
    Pause the workflow and notify an assigned reviewer. They approve, request changes, or reject the draft. The decision and all feedback travel with the run record.
  </Card>

  <Card title="Automated Quality Scoring" icon="star-half-stroke">
    Score each draft against brand alignment, structure quality, and citation readiness before human review is triggered — so reviewers only see drafts that meet a minimum bar.
  </Card>

  <Card title="Conditional Logic" icon="code-branch">
    Branch by quality score, content type, or any block output. High-quality outputs can skip human review; low scores route to a different workflow path or a more intensive review step.
  </Card>
</CardGroup>

## Adding a human review gate

<Steps>
  <Step title="Open a workflow in the builder">
    Navigate to the Workflows section of your workspace and open the workflow you want to add a gate to, or start a new one from scratch.
  </Step>

  <Step title="Drag a Human Review block from the block library">
    Open the block sidebar, go to the **Flow** section, and drag the **Human Review** block onto the canvas.
  </Step>

  <Step title="Connect it between the drafting block and the publishing block">
    Draw a connection from the output of your last generation step to the input of the Human Review block, then connect the Human Review block's output to your publishing block.
  </Step>

  <Step title="Assign an owner">
    Click the Human Review block to open its configuration panel. Select an owner from your workspace members. This person receives a notification when the gate triggers.
  </Step>

  <Step title="Set an SLA (optional)">
    Enter a time window — for example, 24 hours. If the reviewer has not acted within that window, VeluraHQ sends them a reminder. The workflow remains paused until they respond.
  </Step>

  <Step title="Save and deploy the workflow">
    Save your changes. The gate is now active for all future runs. Existing in-progress runs are not affected.
  </Step>
</Steps>

## Reviewing a draft

When a gate triggers, the assigned reviewer receives a notification with a direct link to the run. Inside the review panel, they can:

* **Preview the draft** in the output panel, with each block's output visible in context
* **Compare it against Brand Kit rules** to check voice, tone, and terminology alignment
* Click **Approve** to release the workflow and continue to the next step
* Click **Request changes** and add inline comments — the workflow loops back for revision
* Click **Reject** to halt the run entirely and flag it for follow-up

<Note>
  When a reviewer requests changes, the workflow loops back to the previous generation block with their feedback attached as context. The next generation attempt receives the original inputs plus the reviewer's specific comments, so the model is revising against explicit guidance rather than starting from scratch.
</Note>

## Automated quality scoring

Before a human review gate fires, VeluraHQ can automatically score the draft on three dimensions. Each dimension produces a score from 0 to 100:

* **Brand alignment (0–100)** — How closely the output follows your voice, tone, and terminology rules as defined in your Brand Kit
* **Structure quality (0–100)** — Heading hierarchy, paragraph length, sentence variety, and overall readability
* **Citation readiness (0–100)** — Presence of specific, verifiable claims that AI answer engines can attribute and cite

You set the threshold for each dimension. For example: a score below 70 on brand alignment triggers the human review gate and pauses the workflow; a score of 70 or above routes the draft directly to publish. Thresholds can be set independently for each dimension, so you can require stricter brand compliance while allowing more flexibility on structure.

<Tip>
  On high-volume Sheet runs, set minimum quality thresholds that match your confidence in the workflow's outputs. Requiring a score of 80 or above before auto-publishing means your reviewers only handle the outliers — typically 10–20% of a batch — rather than approving every single row manually. You stay in control of edge cases without creating a bottleneck on clean outputs.
</Tip>

## Bulk review in Sheets

When you run a workflow across a Sheet and some rows trigger a review gate, those rows appear in a dedicated **Review** filter view inside Power Sheets. You do not need to process the entire batch to handle flagged rows.

From the Review view, you can:

* **Approve** individual rows to release them to the next workflow step
* **Request changes** on a specific row and add a comment that feeds back into the next generation
* **Reject** rows that should not continue

Approved rows move through the rest of the workflow immediately. Rows you have not yet acted on remain paused. This lets you work through a review queue at your own pace without holding up the rows that are already cleared.