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

# Questions

> Build and manage the question bank for your NBQ Engine.

The **Questions** tab is where you define every question the NBQ Engine can return. The engine selects from this bank based on the conversation so far and your behaviour configuration.

## Adding questions

<Frame>
  <img src="https://mintcdn.com/zelinqa-d77427f3/FpdEzAeb9Hg8n1by/images/backoffice-questions-categories.jpg?fit=max&auto=format&n=FpdEzAeb9Hg8n1by&q=85&s=a5ad3b032bbc97f633ec9d59849f4c57" alt="Question categories with priority order in the Zelinqa NBQ Studio" width="2048" height="957" data-path="images/backoffice-questions-categories.jpg" />
</Frame>

<Tabs>
  <Tab title="CSV Import">
    Click **Import questions** to load many questions at once:

    1. Download the provided **template**.
    2. Fill in your questions. Only the **questions** column is mandatory. An optional `categorie` column lets you assign categories — empty means the default category, and new categories are created automatically.
    3. Upload the file (CSV, up to 2 MB). Imported questions are **active by default**. Duplicates already present are skipped and reported.
  </Tab>

  <Tab title="One at a time">
    Click **Add Question**, fill in the question text and category, and save. An option lets you keep adding without returning to the list — useful for small sets.
  </Tab>
</Tabs>

## Question fields

| Field                 | Description                                                                                                       |
| --------------------- | ----------------------------------------------------------------------------------------------------------------- |
| **Question text**     | The wording shown to users. The NBQ Engine may refine phrasing at runtime to match your chosen tone.              |
| **Category**          | The group this question belongs to (see below).                                                                   |
| **Active / Inactive** | Only active questions are eligible to be returned by the engine. Deactivate without deleting to pause a question. |

## Categories and priority order

A **category** groups questions that aim to collect the same piece of information — for example, *budget*, *timeline*, or *use case*. By default all categories have equal importance.

You can optionally define a **priority order** by ranking categories top to bottom. The engine then prioritises collecting higher-ranked categories first. You can reorder with the ↑ ↓ controls next to each category, or reset to equal priority at any time.

## Auto-organise with AI

<Frame>
  <img src="https://mintcdn.com/zelinqa-d77427f3/FpdEzAeb9Hg8n1by/images/backoffice-questions-list.png?fit=max&auto=format&n=FpdEzAeb9Hg8n1by&q=85&s=4c4d60ebc25f72eb902eff6a4eb34b67" alt="Questions list in the Zelinqa NBQ Studio" width="2348" height="1218" data-path="images/backoffice-questions-list.png" />
</Frame>

Click **Catégoriser automatiquement** to let NBQ automatically sort your uncategorised questions into coherent groups. Importantly, it also **establishes a priority order** for those categories at the same time — ranking them from highest to lowest so the engine targets higher-priority categories first. Review the result and adjust any category or order as needed. Use **Ajouter une catégorie** to create a group manually, and the **↑ ↓ arrow controls** to move categories up or down and set the priority order as you wish. Use **Réinitialiser l'ordre** to return all categories to equal priority at any time.

<Note>
  Auto-organise only touches questions still in the **default** category. Anything you have already organised is left untouched. It also sets an initial priority order for the categories it creates — you can freely reorder them afterwards using the ↑ ↓ arrows next to each category. It is the fastest way to finish setup right after a CSV import.
</Note>

## How answered questions are tracked

Your application sends the `answered_question_ids` list with every API request. The engine automatically excludes any question whose `external_id` appears in that list — it never repeats a question.

The engine does not store which questions have been answered. Your application owns that state and sends it with each call — this is what makes the API stateless.
