Core Concepts
Knowledge Base
Understanding how your AI's knowledge is structured and managed.
The knowledge base is the foundation of your AI assistant. It contains all the indexed content from your website, organized and optimized for intelligent retrieval.
What's Stored
Page Content
The main text content from each crawled page, cleaned of navigation and boilerplate.
Vector Embeddings
Semantic representations of your content that enable intelligent similarity search.
Metadata
Page titles, URLs, headings, and other structural information for context.
How Retrieval Works
- 1
Query Embedding
The question is converted to a vector embedding
- 2
Similarity Search
Find content chunks with similar meaning (not just keywords)
- 3
Context Assembly
Top matching content is compiled for the AI
- 4
Response Generation
AI formulates a natural answer using the retrieved context
Managing Your Knowledge Base
Re-crawling
When your website content changes, trigger a re-crawl from Studio to update your knowledge base. This can be done manually or happens automatically based on your plan's recrawl frequency.
System Prompt
While the knowledge base provides the "what," the system prompt defines the "how." It guides the AI on tone, behavior, and response style. Edit it in Studio.
Best Practices
- Keep content up to date - Regularly re-crawl to ensure accuracy
- Use clear, structured content - Well-organized pages produce better results
- Cover common questions - Ensure your website has content for frequent queries
- Review chat logs - Identify gaps in your knowledge base from user questions