On this page
AUTOMATION AND AI
Knowledge base (RAG)
A knowledge base, in the context of an AI chatbot, is the collection of business content the bot is allowed to answer from: crawled web pages, pasted articles and FAQ pairs, split into chunks and indexed as embeddings. At reply time the most relevant chunks are retrieved and given to the model as context, a pattern called retrieval-augmented generation (RAG).
embeddingsai agentfallback messageknowledge base sources
Why it matters
Without a knowledge base a language model answers from general training and invents specifics; with one, answers cite your prices, hours and policies. The quality of the base decides the quality of the bot: stale pages give stale answers, and a thin base makes the model guess. Scoping bases per widget or channel keeps a support bot from quoting the sales page.
In VGraple CRM
Sources are added under the chat widget or channel settings as URLs to crawl, pasted text or FAQ pairs; content is chunked at about 512 tokens with overlap, embedded, and re-indexed when a source changes. The five most relevant chunks are passed to the model. The knowledge base guide covers what to include and how to test.
Related terms
Embeddings, AI agent, Fallback message, Conversation summary.