AI knowledge assistant (RAG) over your documentation

An AI knowledge assistant answers questions using your own documentation —manuals, internal policies, catalogs, contracts, historical tickets— instead of the model’s generic knowledge. The technique is called RAG (Retrieval-Augmented Generation) and it’s how AI can speak accurately about your content without making things up and without retraining any model.

What it’s for

• Internal support: let your team ask “what’s the vacation policy?” or “how do I configure X?” and get the exact answer from your documents.
• Onboarding: let new people find answers without interrupting anyone.
• Customer support: answer about your products and terms with up-to-date information.

How to build it

1. Index your documents. With RAG (index) you load your PDFs, pages or text; they are stored as vectors in the knowledge base.

2. Connect the agent. An AI Agent with its Brain and a RAG (retrieve) node hooked to the matching port.

3. Publish the chat. A Trigger Chat and the assistant is available at /c/<token>.

When someone asks, the agent retrieves the fragments most similar to the query and answers with them as context — from the right source, not a generic reply.

Keep the knowledge current

Schedule re-indexing (with a Schedule) so that when your documents change, the knowledge base updates on its own. The assistant always answers with the current version.

Why RAG and not “training a model”

• No retraining: you update knowledge by adding documents, not by retraining (expensive and slow).
• With sources: the agent answers based on a concrete fragment, which reduces hallucinations.
• Your data, private: the information lives in your base, it isn’t “baked” into the model.

Keep reading

• AI agent fundamentals — what RAG is and how it fits with brain, tools and skills.
• Customer support chatbot.
• Sales and lead assistant.