Knowledge Base
Enhance your AI agent with domain knowledge
Knowledge Base
Train your AI agents with domain-specific knowledge to provide more accurate and contextual responses.
Adding Knowledge Sources
ChatterMate supports multiple ways to add knowledge to your AI agents:
PDF Upload
Upload PDF documentation, guides, and manuals
URL Import
Import knowledge from web pages and documentation sites
Link Existing
Share knowledge sources between multiple agents
Knowledge Sharing
ChatterMate allows multiple agents to share common knowledge sources through the “Link Existing” feature:
Link Existing
Connect agents to existing knowledge sources
Shared Access
Multiple agents can reference the same knowledge
Benefits of Knowledge Sharing
- Maintain consistency across different agent types
- Reduce duplicate content processing
- Centralized knowledge management
- Easy updates propagate to all linked agents
Use Cases
- Share company policies across all agents
- Common product documentation for sales and support
- Technical specifications for multiple specialist agents
- FAQ content for general and specialized support
When you update a shared knowledge source, all agents linked to it will automatically have access to the updated information without requiring reprocessing.
Knowledge Processing
Batch Processing
ChatterMate processes knowledge sources in batches to ensure system stability:
- Maximum 20 subpages per batch
- Processing occurs every 5 minutes
- Queued items are processed in FIFO order
- Status tracking for each knowledge source
Knowledge processing includes text extraction, chunking, embedding generation, and vector database indexing. This ensures optimal retrieval during customer interactions.
Processing States
Notifications
When Firebase Cloud Messaging is enabled:
- Notifications are sent only when processing completes
- Success and failure states are reported
- Click notification to view details
- Configurable notification preferences
Knowledge Source Management
Source Types
- PDF Documents
- Web Pages
- HTML Content
- Markdown Files
- Plain Text
Source Details
Each knowledge source shows:
- Source name/URL
- Number of subpages
- Creation date
- Processing status
- Last updated timestamp
Best Practices
-
Document Preparation
- Use clear, well-structured documents
- Break large documents into smaller sections
- Ensure text is extractable from PDFs
- Keep URLs accessible and stable
-
Processing Management
- Monitor queue status
- Process during off-peak hours
- Keep documents under 100MB
- Use consistent formatting
-
Content Organization
- Group related documents
- Use descriptive filenames
- Update outdated content
- Remove obsolete sources
Limitations
- Maximum 20 subpages per batch
- Processing occurs every 5 minutes
- PDF size limit: 100MB
- URL must be publicly accessible
- Text must be extractable (no scanned PDFs)
Monitoring Progress
Track knowledge processing through:
- Dashboard status indicators
- Firebase notifications (if enabled)
- Processing history logs
- Error reports and retry options
What’s Next?
After adding knowledge sources:
- Test AI responses with new knowledge
- Monitor response accuracy
- Update content as needed
- Configure notification preferences
Testing
Next: Learn how to test your AI agent’s knowledge