Use well-structured, high-quality content to improve the accuracy and relevance of AI Agent responses. This article describes the supported content sources, recommended content formats, and best practices for preparing knowledge used by Retrieval-Augmented Generation (RAG).

Who Should Use This Feature?

This information is intended for:

• Administrators configuring content sources for AI Agents.
• Teams creating or managing knowledge content.
• Organizations preparing knowledge repositories for AI-powered responses.

Choose a Content Source

AI Agents support multiple content sources. Select the source that best matches your content management requirements.

Knowledge Base Articles

Use Knowledge Base articles when you need to manage large volumes of structured content.

Knowledge Base articles are recommended when:

• Content is updated frequently.
• Automatic synchronization is required.
• Content is maintained in a centralized knowledge repository.

Documents

Use documents when content is not available in a Knowledge Base.

Document-based content:

• Requires manual upload.
• Must be maintained manually.
• Does not support automatic synchronization.

Q&A Pairs

Use Q&A pairs for static information that requires predefined responses.

Q&A pairs are suitable when:

• Questions and responses are known in advance.
• Responses should be returned as authored.
• Content does not exist in Knowledge Base articles or documents.

Ensure each question represents a specific topic and includes common variations users may ask.

Language Requirements

To maintain consistent retrieval results:

• Use a single language across all content sources within an AI Agent application.
• Use English whenever possible.
• If English is not used, ensure all content is authored in the language supported by the AI Agent.

​

Prepare Knowledge Base Content

Knowledge Base articles are the preferred content source for AI Agents.

Supported Content

Knowledge Base articles support:

• Plain text
• Content blocks
• Content variables
• Tables (feature-controlled)
• Images (feature-controlled)

Unsupported Content

The following content types are not supported:

• Videos
• Audio files
• GIFs
• Embedded documents
• External hyperlinks

​

Best Practices

When creating Knowledge Base content:

• Use clear and descriptive headings.
• Write in complete sentences.
• Organize information into logical sections.
• Keep content concise and easy to understand.
• Avoid fragmented or disconnected content.

Synchronize Knowledge Base Content

Knowledge Base content can be synchronized automatically based on configured filters and sync schedules.

For more information, see Add Content Sources – Knowledge Base.

Prepare Document Content

Use documents for manually maintained content.

Supported Content

Documents support:

• Plain text

Limitations

Consider the following limitations when using documents:

• Large tables may reduce retrieval quality.
• Images are not processed.
• Videos, audio files, GIFs, and embedded documents are not supported.
• Linked content is not indexed.

Best Practices

To improve retrieval quality:

• Use simple formatting.
• Structure content using headings and sections.
• Write clear, complete sentences.
• Avoid complex layouts and heavily formatted content.

Prepare Q&A Content

Use Q&A pairs when responses should be returned exactly as authored.

Create Questions

When creating questions:

• Use clear language.
• Focus on a single topic.
• Include common user variations.

Create Answers

When creating answers:

• Provide accurate information.
• Keep responses concise.
• Avoid combining multiple topics in a single answer.

Control Content Usage with Topic Tags

Use topic tags to control how content is used by AI Agents.

Topic tags help:

• Organize content by subject.
• Improve retrieval relevance.
• Restrict content to specific use cases when required.