Skip to main content
The knowledge base is a searchable repository of information your agents reference when responding to customers. Better knowledge = better responses.

​
What is the Knowledge Base?

Your knowledge base contains all the information your agents can draw from:
  • Product documentation
  • Help articles and FAQs
  • Internal guides and runbooks
  • Troubleshooting procedures
  • Policy documents
When a customer asks a question, agents search this knowledge to find relevant information before responding.

​
Knowledge Sources

Connect external platforms to automatically sync content:
SourceWhat Gets Synced
NotionPages, databases, wikis
ConfluenceSpaces, pages, articles
Google DriveDocs, Sheets, PDFs
ZendeskHelp center articles
IntercomHelp articles
SlackChannel history
Web/URLsAny public webpage
Content syncs automatically and stays up to date as you make changes in the source.

​
Custom Knowledge

Create articles directly in Duckie when you need knowledge that doesn’t exist elsewhere:
  • Quick answers to common questions
  • Internal procedures not documented elsewhere
  • Temporary information (promotions, outages)
  • Corrections or clarifications

​
Knowledge Gaps

When agents can’t confidently answer a question, it’s logged as a knowledge gap:
  • See what questions aren’t being answered
  • Track how often each gap occurs
  • One-click to create an article that fills the gap
This creates a feedback loop: customer questions → identify gaps → create knowledge → better answers.

​
Knowledge Tags

Organize and control access with tags:
  • Tag articles by topic, product, team, or any other dimension
  • Filter agent access so agents only search relevant knowledge
Example:
  • Billing Agent → Only sees articles tagged “billing”, “payments”
  • Technical Agent → Only sees articles tagged “technical”, “troubleshooting”

​
How Search Works

When an agent needs information:
  1. Query understanding — Analyzes the customer’s question
  2. Semantic search — Finds articles with similar meaning (not just keywords)
  3. Relevance ranking — Orders results by relevance
  4. Context extraction — Pulls the most useful parts
The agent then uses this context to formulate an accurate response.
Semantic search means agents find relevant information even when customers use different words than your documentation. “Can’t log in” matches articles about “authentication issues” and “password reset.”

​
Getting Started

1

Connect a Source

Connect your primary documentation source (Notion, Confluence, etc.).
2

Wait for Sync

Let Duckie index your content (usually 2-5 minutes).
3

Test in Playground

Ask questions to verify the agent finds relevant information.
4

Fill Gaps

Review the Gaps tab and create articles for unanswered questions.

​
Next Steps

Connect Sources

Sync external documentation

Custom Knowledge

Create articles in Duckie

Knowledge Gaps

Fill gaps from customer questions

Knowledge Tags

Organize and filter knowledge