kibana/docs/reference/configuration-reference/ai-assistant-settings.md

navigation_title	mapped_pages	applies_to
AI Assistant settings	https://www.elastic.co/guide/en/kibana/current/ai-assistant-settings-kb.html	deployment self all

AI Assistant settings in {{kib}} [ai-assistant-settings-kb]

xpack.productDocBase.artifactRepositoryUrl

Url of the repository to use to download and install the Elastic product documentation artifacts for the AI assistants. Defaults to https://kibana-knowledge-base-artifacts.elastic.co

Configuring product documentation for air-gapped environments [configuring-product-doc-for-airgap]

Installing product documentation requires network access to its artifact repository. For air-gapped environments, or environments where remote network traffic is blocked or filtered, the artifact repository must be manually deployed somewhere accessible by the Kibana deployment.

Deploying a custom product documentation repository can be done in 2 ways: using a S3 bucket, or using a CDN.

Deploying using a S3 bucket [_deploying_using_a_s3_bucket]

1. Download the artifacts for your current {{kib}} version

The artifact names follow this pattern: kb-product-doc-{{productName}}-{{versionMajor}}.{{versionMinor}}.zip

The available products are:

• elasticsearch
• kibana
• observability
• security

You must download, from the source repository (https://kibana-knowledge-base-artifacts.elastic.co/), the artifacts for your current version of Kibana.

For example, for Kibana 8.16:

• kb-product-doc-elasticsearch-8.16.zip
• kb-product-doc-kibana-8.16.zip
• kb-product-doc-observability-8.16.zip
• kb-product-doc-security-8.16.zip

2. Upload the artifacts to your local S3 bucket

Upload the artifact files to your custom S3 bucket, then make sure that they are properly listed in the bucket’s index, similar to the bucket listing displayed when accessing https://kibana-knowledge-base-artifacts.elastic.co/ in a browser.

3. Configure {{kib}} to use the custom repository

Add the following line to your {{kib}} configuration file:

# Replace with the root of your custom bucket
xpack.productDocBase.artifactRepositoryUrl: "https://my-custom-repository.example.com"

4. Restart {{kib}}

You should then be able to install the product documentation feature from the AI assistant management page.

Deploying using a CDN [_deploying_using_a_cdn]

Deploying using a CDN is quite similar to the S3 bucket approach. The main difference will be that we will need to manually generate the bucket listing and set it as the CDN folder’s index page.

1. Download the artifacts for your current {{kib}} version

Following the step from the Deploying using a S3 bucket section

2. Upload the artifacts to the CDN

Create a folder in your CDN, and upload the artifacts to it.

3. Create and upload the bucket listing

Generate the S3 bucket listing xml file for the folder.

To do that, copy the following template, and replace the versions in the <Key> tags with your current version of {{kib}}.

For example for {{kib}} 8.17, replace all 8.16 occurrences in the file with 8.17.

<ListBucketResult>
    <Name>kibana-ai-assistant-kb-artifacts</Name>
    <IsTruncated>false</IsTruncated>
    <Contents>
        <Key>kb-product-doc-elasticsearch-8.16.zip</Key>
    </Contents>
    <Contents>
        <Key>kb-product-doc-kibana-8.16.zip</Key>
    </Contents>
    <Contents>
        <Key>kb-product-doc-observability-8.16.zip</Key>
    </Contents>
    <Contents>
        <Key>kb-product-doc-security-8.16.zip</Key>
    </Contents>
</ListBucketResult>

Then upload that xml file to the same CDN folder where the artifacts were uploaded, and then configure the folder to have that file served as the folder’s index.

4. Configure {{kib}} to use the custom repository

Add the following line to your {{kib}} configuration file:

# Replace with the path to the CDN folder previously configured
xpack.productDocBase.artifactRepositoryUrl: "https://my-custom-repository.example.com"

5. Restart {{kib}}

You should then be able to install the product documentation feature from the AI assistant management page.