LogoLogo
  • GETTING STARTED
    • Welcome to the NLX platform
    • How NLX works
    • Guides
      • Analytics dashboard
      • Chat
      • Generative Journey (Slots)
      • Model Context Protocol
      • Voice
      • Voice+ script
      • Touchpoint components
        • Carousel modality
        • Video modality
    • Terms & concepts
    • Generative AI
    • Developer
  • Build
    • Workspace setup
    • Flows
      • Intro to flows & variables
      • The Canvas
      • Flow settings
      • Nodes
      • Flow appearance
    • Resources
      • Actions
        • Implementation
        • Request model
      • Analytics tags
      • Context variables
      • Data requests
        • Implementation
        • Response model
        • Request model
      • Knowledge bases
        • Ingest content
        • Add metadata
        • Support citations
        • Employ KB
      • Lifecycle hooks
        • Implementation
      • Modalities
      • Secrets
      • Slots (custom)
        • Adding values
        • Translating slots
      • Voice+ scripts
        • Add + download script
        • Deploy script + install SDK
        • Create Voice+ flow
    • Integrations
      • Channels
        • Alexa
        • Amazon Chime SDK
        • Amazon Connect
        • AWS End User Messaging SMS
        • AWS End User Messaging Social
        • Bandwidth
        • Genesys
        • Twilio
        • Zendesk Sunshine
      • LLM services
        • Amazon Bedrock
        • Anthropic
        • Azure OpenAI
        • Cerebras
        • Cohere
        • Google Vertex AI
        • Groq
        • NVIDIA
        • OpenAI
        • xAI
      • NLP
        • Amazon Lex
        • Google Dialogflow
        • Google Dialogflow CX
        • Custom NLP
    • Translations
  • Deploy & test
    • Applications
      • Attach flows
      • Assign default behavior
      • Add channels
        • API
          • REST API
        • Alexa
        • Amazon Chime SDK
        • Amazon Connect
        • AWS End User Messaging SMS
        • AWS End User Messaging Social
        • Genesys
        • Genesys SMS
        • Messenger
        • Microsoft Teams
        • Slack
        • SMS via Bandwidth
        • Twilio SMS
        • Twilio Voice
        • WhatsApp via Twilio
        • Zendesk Sunshine
      • Deploy
      • Optional: Set lifecycle
      • Optional: Set languages
    • Test
      • Test a conversation
      • Automated tests
      • Test an external integration
  • Analyze
    • Conversations
    • Analytics
      • Creating dashboards
      • Formulas & multi-metrics
      • Canvas analytics
    • Training
  • Workspace Settings
    • Escalation channels
    • Resource tags
    • Audit
  • Admin
    • Access control
      • Roles & permissions
    • Notifications
    • FAQ
    • Contact support
Powered by GitBook
On this page
  • Option 1: Assign to application
  • Option 2: Invoke in flow/Custom flow
  1. Build
  2. Resources
  3. Knowledge bases

Employ KB

Learn how to put your knowledge base to work in NLX

Last updated 14 days ago

Knowledge base content is useful for responding to user utterances that would not be a match to a flow but are common questions or requests for information that you want to support.

You may choose to assign your knowledge base as the default process whenever an utterance is logged as "unknown" (i.e., unmatched to a flow), or strategically invoked in a flow when an utterance results in a "No match" on any user-type node.


Option 1: Assign to application

For any user utterance that your conversational AI cannot understand, the query will be checked against the Knowledge base 's Unknown default. Best for simple, fast integrations of knowledge base content.

For more advanced setup that supports metadata, citations, and follow-up responses, check .

User queries that cannot be matched to content in your Knowledge base triggers the application's .

To employ your knowledge base through the Unknown default behavior:

  • Navigate to an application in your workspace > Click the application's Default behavior tab

  • Under Unknown behavior, toggle from Quick replies to Knowledge base

  • Select your Knowledge base from the dropdown

  • Click Save

  • Create a new application build and deploy through the application's Deployment tab to experience your new knowledge base


Option 2: Invoke in flow/Custom flow

User queries that cannot be matched to content in your Knowledge base eject out of the Knowledge base node's No match edge.

To employ your knowledge base through a custom process:

  • Navigate to Flows in the workspace menu > Select a flow

  • Assign a Knowledge base implemented in the workspace to the node

  • Link a Basic node from the Knowledge base node's Match edge > Enter the placeholder name given and the answer {"yourKBname".answer} as a message on the Basic node

  • Proceed from the Basic node to the rest of your flow or redirect to another flow

  • Click Save

At key points in a conversation flow(s), you may decide to include a Knowledge base node to handle unmatched queries. You may also create a custom flow that checks one or more Knowledge bases when your application's is triggered. Whether you decide to use a single centralized flow or decide to place Knowledge base nodes throughout different flows in your application, use of the Knowledge base node supports metadata filtering and in-line citations in its responses.

Add a to the Canvas > Provide a name for reference locally in the flow

If you have provided to your Knowledge base, you may enter IF comparator statements = to provide a filter(s) whenever the node is hit

If you want to support citations, follow

Create an application build with and deploy to make your knowledge base live to users.

Unknown behavior
Knowledge base node
Metadata
setup here
your flow
assigned to your application
Option 2
Fallback response