Knowledge Base

Knowledge bases are collections of information that your AI agents use to answer customer questions. You can add documents, websites, and text content to train your agents on your specific domain.

Creating a Knowledge Base

1. Go to Knowledge Base in your workspace sidebar
2. Click "Create"
3. Give it a descriptive name (e.g., "Product Documentation")
4. Add your content sources

Content Sources

File Upload

Upload documents directly to your knowledge base:

• PDF — documentation, guides, manuals
• DOCX — Word documents
• TXT — plain text files
• CSV — structured data

Website Crawling

Import content from your website or documentation:

1. Enter the starting URL (e.g., https://docs.example.com)
2. The crawler follows links within the same domain
3. Content is extracted and indexed automatically

The crawler respects robots.txt and only processes publicly accessible pages.

Text Sources

Paste or type content directly — ideal for FAQs, policies, and quick reference material.

Folder Organization

Knowledge bases are organized as folders containing multiple sources. This helps you:

• Group related content together (e.g., "API Docs", "FAQs", "Policies")
• Manage indexing per folder
• Attach specific folders to different agents

Indexing

After adding content, indexing makes it searchable by your agents:

1. Content is extracted from documents and web pages
2. Text is split into searchable chunks
3. Vector embeddings are generated for semantic search
4. Vectors are stored in a searchable database

Indexing Status

Initial provisioning can take a few minutes. Subsequent indexing is faster.

Updating Content

1. Add, remove, or modify sources in your knowledge base
2. Click "Re-index" to update
3. Wait for indexing to complete

Agents automatically use the latest indexed content — no need to reconfigure them.

Best Practices

• Use well-structured, clearly written documents for best results
• Keep content focused and relevant to your use case
• Organize sources into logical folders
• Remove outdated information to improve accuracy
• Re-index after any content changes
• Test knowledge retrieval in the Playground before deploying