Managing knowledge sources

Manually import multiple knowledge bases (KBs) using CSV files, so AI agents generate responses that are tailored to your organization. Each uploaded file is treated as an independent knowledge source and can be classified by the system based on its intended audience, such as technical support or employees. This classification determines how AI agents use the content during search and resolution, including whether it can be exposed to employees or used internally for reasoning.

Spark can use knowledge sources from a wide range of platforms. This documentation uses ServiceNow as an example.

Understanding the knowledge sources list

The list shows all imported knowledge sources with the following details:

• Name: The filename of the uploaded CSV.

• Origin: The source of the article; always marked as Custom.

• Target group: The intended audience of the knowledge source. AI agents use this attribute during search and ranking to surface the most relevant content based on context:

  • Employee: end-user-facing knowledge Spark can expose to the employee.

  • Technical support: IT or support-facing knowledge Spark uses for troubleshooting.

• Size: The number of articles in the CSV. The system can also display the following statuses:

  • Failed: If the file cannot be processed, for example, if it contains a syntax error. In this case, hover over the error message to see more details and a remediation proposal.

  • Processing: If the file contains many articles, it may take the system time to process.

• Type: The file type of the knowledge source. Currently, the system supports CSV only.

• Last update: The timestamp of when the knowledge source was last updated. It helps to assess if an article is up to date.

Using the action menu for knowledge sources

Select an action in the list, and click the action menu on the right side of the row to view the available options:

• Export: Download the original CSV file corresponding to the last successful import. Each knowledge source can be exported individually only.

• Delete: Permanently remove a previously uploaded CSV and all knowledge articles associated with that knowledge source.

Uploading knowledge base articles

The following procedure uses ServiceNow as an example; however, on a high level, the process is similar in other ITSM platforms.

1

Identify data to export from your ITSM

1. Open the list of knowledge base articles in ServiceNow.

2. Select knowledge base articles by applying these filters:

   • Published articles only.

   • Articles available to your service desk users. Spark indexes every uploaded article regardless of ITSM permissions.

   • Exclude translated versions of the same article.

   • Include every article you want to import.

Manual uploads are not incremental. Plan to export all the data required for the Spark knowledge base. When uploading a new file, the new content replaces the previous content.

2

Export data from your ITSM

Export data from ServiceNow by performing a standard CSV export:

1. Navigate to your knowledge base articles list.

2. Click Apply filter to configure the desired filters.

3. Click the cogwheel icon > Personalize list.

4. Include the following columns in the CSV, using the exact case-sensitive column names:

Column name	Description
number	Identifier of the knowledge base article. Example: KB0012345
short_description	Title of the knowledge base article. Example: How to troubleshoot network issues
kb_knowledge_base	Name of the knowledge base the article belongs to Example: IT Knowledge
kb_category (optional)	Category of the knowledge base article. Example: Network
sys_updated_on (optional)	Last update date. Example: 11-06-2025
text	Article content, in one of the supported formats: plain-text, markdown or HTML. Example: <p>test kb</p>

1. Click the action menu in any column header and select Export > CSV to download the file.

Ensure the file is UTF‑8 encoded. If your ServiceNow instance is not configured to perform UTF-8 exports by default, convert the file to UTF-8 afterward.

3

Upload articles into Nexthink

1. Go to Spark  > Manage knowledge sources.

2. Select Import in the upper right corner.

3. Enter a full, parameterized URL as the External link that links to individual knowledge articles in your knowledge system. The URL must include a placeholder for the article identifier so that AI agents can generate direct links to the original content. Examples:

   • ServiceNow: https://<instance>.service-now.com/sp?id=kb_article&sysparm_article={number}

   • BMC Helix: https://<instance>.onbmc.com/portal/kb/article/{number}

   • JSM: https://<instance>.atlassian.net/servicedesk/customer/kb/view/{number}

   

Root-only hostnames are not supported.

1. Select a CSV file of up to 100 MB from your local device to import it into the system.

2. Define a Target group for the knowledge source:

   • Technical support: Users responsible for diagnosing and resolving IT issues.

   • Employee: End users consuming IT services for their daily work.

3. Select Done, and the system takes the following steps:

   • If you have already manually uploaded a file, the old content will be discarded and replaced by the new content.

   • File processing occurs asynchronously and may take up to 15 minutes for large knowledge bases.

Updating an existing knowledge source

To ensure Spark always has a knowledge base available and can answer questions even when the system is processing a new version of a knowledge source file, Nexthink recommends the following order:

1. Upload the updated knowledge source file while still leaving the old one in the system.

2. Once the new knowledge source file is processed, delete the old file.

Best practices

Nexthink recommends the following to improve the selection of knowledge base articles:

• Keep articles current and up to date.

• Use HTML formatting for optimal results.

• Include key information in the article body text.