Skip to main content

Knowledge Bases

Knowledge bases let you give your AI agents access to your organization's specific information. By uploading documents and crawling websites, you create searchable collections that agents can query during conversations—enabling them to answer questions with context from your actual content.

How Knowledge Bases Work​

When you add content to a knowledge base, Gen8 processes it for semantic search. This means the AI can find relevant information based on meaning, not just exact keyword matches. Ask about "customer refund policies" and the AI finds relevant content even if those exact words don't appear.

Each knowledge base is a separate collection you can connect to specific agents. This lets you create focused knowledge bases for different purposes—one for product documentation, another for HR policies, another for technical specifications.

Creating a Knowledge Base​

Navigate to Organization → Knowledge Bases and click Create Knowledge Base. Give it a descriptive name and a brief description of what it will contain. Good names make it easier to choose the right knowledge base when configuring agents later.

Adding Content​

Uploading Files​

The most direct way to add content is uploading files. Click Upload Files, then drag and drop your documents or select them from your computer. Gen8 processes each file and indexes it for search.

Supported Formats​

CategoryFormats
DocumentsDOC, DOCX, PDF, TXT, MD
PresentationsPPTX
CodeC, CPP, CS, CSS, GO, HTML, JAVA, JS, JSON, PHP, PY, RB, SH, TEX, TS

Importing from Websites​

For content that lives on the web, you can configure automatic crawling. Click Add Website Source, enter the URL, and set up your crawl preferences. Gen8 will fetch the content and keep it updated based on your schedule.

See Website Crawling for complete details on configuring crawl behavior, frequency, and URL exclusions.

Using Knowledge Bases with Agents​

Once you have content in a knowledge base, connect it to your agents:

  1. Open your agent in the editor
  2. Go to the Tools section and add Search Knowledge Base
  3. Select which knowledge base this agent should search
  4. Save the agent

Now when users ask questions, the agent can search your documents to find relevant information and incorporate it into responses.

Testing Your Knowledge Base​

Before putting an agent into production, test how well your knowledge base handles expected queries. Open the knowledge base and use the Test Search feature to run sample queries.

Good queries to test:

  • Questions users actually ask
  • Variations of common questions
  • Edge cases that might be tricky

Review the results to see which documents surface and how relevant the excerpts are. If results aren't satisfying, you might need to add more content, restructure existing documents, or create a more focused knowledge base.

Managing Content​

Monitoring Status​

Each file in your knowledge base shows its processing status:

StatusMeaning
ProcessingBeing indexed for search (takes a few moments)
CompletedReady for agent searches
FailedSomething went wrong—check the file format and try again

Keeping Content Fresh​

Knowledge bases are most useful when their content stays current. Remove outdated files that might lead to incorrect answers, upload new versions when documents are updated, and review your website crawl settings to ensure external content stays synchronized.

Organizing Your Knowledge Bases​

Create Focused Collections​

Rather than one massive knowledge base with everything, consider creating purpose-specific collections. A knowledge base focused on product documentation will give more relevant results for product questions than one that also includes HR policies, sales materials, and technical specs.

Use Descriptive Names and Descriptions​

When you have multiple knowledge bases, clear naming helps you (and other admins) choose the right one. Include the scope in the name: "Customer Support - Product FAQs" is more helpful than "KB-3".

Document What Each Contains​

The description field isn't just metadata—it helps you remember what's in each knowledge base months later. Note the types of documents, the sources, and what questions this knowledge base is meant to answer.

Best Practices​

Structure content for searchability. Documents with clear headings, well-organized sections, and explicit topic statements index better than walls of unstructured text.

Test with real questions. The best way to know if your knowledge base works is to ask it questions your users actually ask. Adjust content based on what does and doesn't surface correctly.

Keep documents focused. A 5-page document about one topic often indexes better than a 50-page document covering many topics. Consider breaking large documents into logical sections.

Review and prune regularly. Outdated content doesn't just take up space—it can lead to incorrect or confusing answers. Schedule regular reviews to keep your knowledge bases accurate.

Next Steps​