Knowledge Base

Add knowledge from URLs, documents, and text to power AI-driven customer responses.

Overview

The Knowledge Base node lets you provide your AI assistant with information from multiple sources - websites, uploaded documents, and custom text. When connected to a Query Processor, your assistant can answer customer questions using this knowledge, providing accurate and helpful responses based on your content.

Use this node when you need to give your AI access to documentation, policies, product information, or any other content that helps answer customer questions.

Configuration

Knowledge Sources

Type: list Required: Yes (at least one source)

The Knowledge Sources list contains all the content your AI can reference. You can add three types of sources:

URL Sources

Add a web page URL and the system will extract its content for your AI to reference.

Fields:

• Name - A descriptive label for this source (e.g., “Product Documentation”)
• URL - The full web address (https://example.com/page )

Notes:

• URLs are automatically normalized (https:// is added if missing)
• The page must be publicly accessible (no login required)
• Content is extracted and indexed for AI use

File Sources

Upload documents directly for your AI to reference.

Supported Formats:

• PDF (.pdf)
• Word Documents (.docx)
• Plain Text (.txt)
• HTML (.html)
• EPUB (.epub)

Limits:

• Maximum file size: 20MB per file

Fields:

• Name - A descriptive label for this document

Text Sources

Add custom text content directly without needing an external source.

Fields:

• Name - A descriptive label for this content
• Content - Your text content (supports multi-line text)

Best for:

• Company-specific information not available elsewhere
• Custom instructions or policies
• Quick additions without creating files

How It Works

1. Add Knowledge Sources - Click “Add” to add URLs, upload files, or enter text content
2. Content is Processed - Your sources are processed and indexed automatically
3. AI Uses Knowledge - When connected to a Query Processor, the AI references this knowledge to answer customer questions
4. Automatic Updates - URL sources can be refreshed to capture content changes

Connection Rules

Important: Knowledge is only available to Query Processor nodes that are directly connected to the Knowledge Base node. This ensures content isolation between different parts of your flow.

Examples

Basic Example: Company Website FAQ

Add your company’s FAQ page as a knowledge source:

Configuration:

• Click “Add URL”
• Name: Company FAQ
• URL: https://yourcompany.com/faq

The AI can now answer questions based on your FAQ page content.

Advanced Example: Multi-Source Knowledge Base

Combine multiple sources for comprehensive coverage:

Sources:

1. URL Source - Product documentation website

   • Name: Product Docs
   • URL: https://docs.yourcompany.com/products

2. File Source - Pricing PDF

   • Name: Pricing Guide 2025
   • Upload: pricing-guide-2025.pdf

3. Text Source - Business hours and contact

   • Name: Contact Information
   • Content:

     Business Hours:
     Monday - Friday: 9 AM - 6 PM EST
     Saturday: 10 AM - 4 PM EST
     Sunday: Closed
     
     Support Email: support@yourcompany.com
     Sales Phone: (555) 123-4567

This setup allows your AI to answer questions about products, pricing, and contact information.

Best Practices

• Use specific URLs - Link to specific pages rather than entire websites. /products/widget-pro is better than just /products
• Keep content focused - Shorter, focused content yields more accurate AI responses
• Organize by topic - Create separate Knowledge Base nodes for different knowledge domains if needed
• Name sources clearly - Use descriptive names like “Return Policy” instead of “Doc1”
• Update regularly - Keep your content current, especially for time-sensitive information
• Combine with FAQ Loader - Use both nodes together for comprehensive coverage (FAQ Loader for structured Q&A, Knowledge Base for documents and web content)

Common Issues

”URL content not appearing in responses”

Solutions:

• Verify the URL is publicly accessible (not behind a login)
• Check that the page contains meaningful text content
• Ensure the Knowledge Base node is connected to a Query Processor
• Try accessing the URL in an incognito browser window to confirm public access

”File upload failed”

Solutions:

• Verify the file is under 20MB
• Check that the file format is supported (PDF, DOCX, TXT, HTML, EPUB)
• Ensure the file is not corrupted - try opening it locally first
• Try re-saving the document in a different supported format

”AI not using the knowledge I added”

Solutions:

• Confirm the Knowledge Base node is connected to your Query Processor
• Check that all sources show as synced (not pending or error)
• Ensure your question relates to the content you added
• Try asking a more specific question that clearly relates to your content

”Source stuck on ‘pending’”

Solutions:

• Wait a few moments - processing can take time for large files
• Refresh the page and check again
• Try removing and re-adding the source
• Check your internet connection

Related Nodes

• FAQ Loader - Use for structured question/answer pairs. Best when you have explicit Q&A content. Knowledge Base is better for documents and web content.
• Query Processor - Receives knowledge from this node and uses AI to answer customer questions based on the content you provide.
• Start Node - Begins the call flow before routing to your knowledge-powered Query Processor.

FAQ Loader vs Knowledge Base

Tip: Use both together for best results - FAQ Loader for your most common questions, Knowledge Base for supporting documentation.