What is a Knowledge Base?

What is it?

A Knowledge Base is a structured collection of information that a system can search and use to answer questions. In RAG (Retrieval-Augmented Generation), it is the source of truth — not the language model's general training data.

Think of it as a library your AI can consult: manuals, policies, product sheets, support articles, or any documents you choose to include.

Why does it exist?

Large language models know a lot about the world, but they do not automatically know your company's procedures, products, or internal documents. A Knowledge Base gives the system access to that specific information at answer time.

Without a Knowledge Base, an AI might guess, hallucinate, or give generic answers. With one, answers can be grounded in documents you control.

How does it work conceptually?

  1. You add documents (PDFs, text files, and other supported formats).
  2. The system extracts text, splits it into chunks, and indexes it for retrieval.
  3. When a user asks a question, the system finds relevant chunks from the Knowledge Base before generating an answer.
Documents → Extraction → Chunks → Index → Available for retrieval

Example

A company stores equipment manuals in its Knowledge Base. A technician asks:

"What maintenance is required after 500 operating hours?"

The system searches indexed manual content — not random web knowledge — to find the relevant section.

Where does it fit in a RAG pipeline?

The Knowledge Base is the input layer of RAG:

Knowledge Base (your documents) ↓ Preparation (extract, chunk, embed) ↓ Vector index (searchable storage) ↓ Retrieval (find relevant chunks) ↓ Generation (LLM answer with context)

Search finds information in the Knowledge Base. RAG retrieves from it and then generates an answer. Chat is often a conversational interface that uses RAG underneath.

How this relates to Noulek

In Noulek, Knowledge Base is the WordPress admin area where you upload and manage documents. Noulek does not automatically index all WordPress posts — you upload or ingest content you want in Search and Chat.

  • Documents appear in Knowledge Base → Documents list with statuses like Processing and Ready.
  • Only Ready documents are available for retrieval.
  • Search and Chat query this indexed content when visitors ask questions.

For practical steps, see How to upload a document.