Configure AI Assistant

To make the AI assistant available to users in your organization, as an administrator you must complete the following:

1. Connect at least one large language model (LLM) provider to your GoodData organization, see Add LLM Providers.

2. Grant workspace users the permission to use the assistant, see Assign Permissions.

Additionally, you can:

• Override the default provider on selected workspaces, see Use Multiple LLMs.

• Control how many assistant queries each user can run within a time window for a given workspace, see Manage User Quotas.

Add LLM Providers

GoodData relies on an external LLM provider to generate responses. For each provider you add, supply the provider’s API key.

Steps:

1. In Settings, go to AI > LLM providers and click Manage:

   
   

2. Click Create:

   
   

3. Enter the LLM provider display name and API key, then Save:

   
   

   The provider is set up:

   
   

The first provider you add becomes the default for every workspace. It may take up to a minute for the provider to propagate to every workspace. To assign a different provider to a specific workspace, see Use Multiple LLMs.

Configure Azure OpenAI for AI Assistant

If you are using Azure OpenAI instead of OpenAI’s public API, ensure that your Azure deployment is configured with the correct model version and settings. The AI Assistant will not function properly unless the Azure deployment meets these requirements:

Assign Permissions

By default, no users have a permission to access the assistant. To enable access, grant users the Workspace.AI_USE_ASSISTANT permission.

Steps:

1. In Users & groups next to a user or group, click … and select Add workspace permission:

   
   

2. Select a workspace, enable AI Assistant, and click Add:

   
   

Use Multiple LLMs

By default the first LLM provider you add becomes the default LLM provider for all workspaces. If you add more than one LLM provider, you can choose which one to use for each workspace in the workspace settings.

Steps:

1. Go to Workspaces.

2. Hover over your workspace and open the … menu. Select Detail.

   
   

3. Switch to Settings and under AI > LLM providers click Manage.

   
   

4. Select which LLM provider you want to use and click Apply.

   
   

Manage User Quotas

GoodData caps assistant usage separately per user and per workspace. By default, each user may submit up to 30 interactions during any rolling 24‑hour period in a given workspace. Once that cap is reached, the user is throttled to one interaction per hour in that workspace until the 24‑hour window elapses.

Use the following API call to change the maximum number of interactions, or the reset period, for a specific workspace:

curl $HOST_URL/api/v1/entities/workspaces/<workspace_id>/workspaceSettings \
    -X POST \
    -H "Authorization: Bearer $API_TOKEN" \
    -H "Content-Type: application/vnd.gooddata.api+json" \
    -d '{
        "data": {
            "attributes": {
                "type": "AI_RATE_LIMIT",
                "content": {
                    "maxInteractions": <number_of_interactions>,
                    "timeWindowHours": <number_of_hours>
                }
            },
            "id": "ai_rate_limit1",
            "type": "workspaceSetting"
        }
    }'

Control AI Visibility of Objects

Some analytical objects are created for technical or internal purposes and should not be used by AI when generating insights or answering questions. To support this use case, GoodData provides a way to control whether an object is visible to AI.

What the isHidden Property Does

The isHidden property controls whether an analytical object is available to AI-powered features, such as:

• AI Assistant search results
• Object suggestions and autocomplete
• Ad-hoc visualizations created by AI

When an object is hidden, AI does not use it or suggest it, even though the object still exists in the workspace. The object remains part of the metadata. It is still available in system tools such as the Analytics Catalog. Also, it can still be used manually in dashboards and visualizations.

Supported Object Types

You can control AI visibility for the following object types:

• Metrics
• Attributes
• Labels
• Facts
• Visualization objects

Default Behavior

The isHidden property is optional.

• If the property is not set (default) or set to false, the object is visible to AI
• If isHidden is set to true, the object is hidden from AI

This ensures full backward compatibility. Existing workspaces behave exactly the same unless you explicitly change the setting.

Example:

{
  "id": "<object-id>",
  "type": "metric",
  "title": "<object-title>",
  "description": "<object-description>",
  "isHidden": true,
  "workspaceId": "<workspace-id>"
}

Behavior with Attributes and Labels

If you hide an attribute, its labels are typically also hidden from AI to keep the experience consistent. This helps prevent AI from using technical labels that belong to hidden attributes.