Without RAG, AI chatbots hallucinate — they generate plausible but incorrect answers. Retrieval-augmented generation fixes this by searching your indexed documents before the AI composes a response.
How RAG works
When a customer asks a question, the system embeds the query, searches your knowledge base for relevant chunks, and includes those passages in the AI's context window. The model then generates an answer grounded in your actual content.
What to index
Product docs, pricing pages, FAQs, policy documents, API references, and internal playbooks. Keep content updated — stale knowledge bases lead to outdated answers.
OmniCrab's approach
Upload PDFs, Markdown files, and URLs directly. OmniCrab handles chunking and retrieval automatically — use the Test Chat tab on a bot to try real questions and check how it answers before publishing.
