An AI knowledge base is a collection of your own content that an AI agent searches during a conversation and uses as context for its reply. Prompts are hard to maintain and easy to outgrow. A knowledge base lets your agent draw on policies, catalogs and help articles that would never fit in a prompt, and you update the content without rewriting the agent.
When you create a knowledge base, Retell splits each source into chunks, embeds them and stores them in a vector database. During a call, before each reply, it condenses the recent conversation into a short search query, pulls the closest chunks (3 by default) and adds them to the LLM prompt. No prompt changes are needed, and an agent can use more than one knowledge base.



