sus-tech

RAG Knowledge Base

Retrieval-augmented generation turns PDFs, wikis and SharePoint libraries into a knowledge base you can ask questions. The model answers from your documents and links the passage it used, so people can check every answer. We build the pipeline in n8n: documents are read, split, indexed and kept up to date.

RAG Knowledge Base

What you get

  • A chat for your employees that answers from your own documents
  • A source for every answer, linked to the document and the passage
  • Updates without extra work: new and changed documents are indexed automatically
  • Answers only from documents a person is allowed to read

How it works

  1. n8n watches the folders, SharePoint libraries or wiki spaces you choose and reads new documents.
  2. The text is split into passages and stored as embeddings in a vector database such as Qdrant or PostgreSQL with pgvector. A local model can compute the embeddings, so the documents never leave your network.
  3. A question retrieves the passages that fit best. The language model answers from them, a cloud model or one on your own hardware served with vLLM or Ollama.
  4. The answer comes with its sources, in a chat or inside a tool your team already uses.

Where it helps

  • Service desks that answer the same questions from long manuals
  • Policies and compliance documents that nobody finds when they need them
  • Onboarding of new employees

Our post Internal knowledge bases with RAG explains the architecture in detail. If you are not sure whether RAG or fine-tuning fits your case, read RAG vs. fine-tuning.

Ready to get started?

Tell us which process you want to improve. We look at it with you and suggest how to start.

Request consultingMore solutions