Skip to main content

Create a Knowledge base

A Knowledge is the store of resources your AI searches when answering questions. This guide takes you from an empty Knowledge to a configured one you can chat with.

📖WHAT IS A KNOWLEDGE?

A Knowledge holds your resources (documents, FAQs, manuals) and the rules for processing them. It is what Rational AI's RAG system searches to ground its answers in your own data.

Prerequisites​

Before you start, make sure you have:

  • Administrative access: an account with Knowledge management privileges.
  • A list of your data sources: the documents, FAQs, or manuals that will feed the Knowledge (see Settings → Sources).
  • The required extensions installed:
    • Time Server: needed for time-based searches (filtering by date or temporal relevance).
    • Rational Knowledge - RAG Query Server: enables advanced queries and the RAG model on the Knowledge.
  • AI models configured: at least one Text Generation model and one Embedding model, active in the AI Model Registry (see Settings → AI Model Registry).

Add a Knowledge​

  1. Go to Settings → Knowledge and select + New.
  2. Enter a clear name (for example Data Marketing 2026 KB).
  3. Add a description of the content and purpose. This helps your team maintain it later.

The AI Search tab controls how resources are prepared for retrieval. Configure these in order:

  1. Turn on Allow AI search. This enables the RAG process for the Knowledge.
  2. Select an Embedding model. This model converts text into vectors so the system can measure how similar a resource is to a query.
  3. Choose a Chunking type. Chunking splits each resource into smaller pieces before embedding. Semantic Chunker is a good default.
  4. Select Save.
📝EMBEDDING WORKS ON CHUNKS

Embedding runs on the small chunks of a resource, not the whole text. The resulting vectors are what RAG compares against a user's question.

Default Touchpoint​

Creating a Knowledge also creates a Touchpoint with default settings, including a working Graph RAG tool. Customize it in Settings → Touchpoints, where you can set the conversation model and the system prompt for this Knowledge.

By default, the AI agent answers in a set order:

  1. Analyze data structure: the agent examines the schema of the data in the Knowledge.
  2. SQL query: if a database can answer the query, the agent runs a SQL query and uses the result.
  3. Graph RAG: if no database has the answer, the agent uses the Graph RAG tool to search the whole Knowledge. To tune how Graph RAG retrieves, see .

Start a chat with your Knowledge​

Your Knowledge is created but still empty. You can open a chat against it now and add data as you go.

  1. Open the Knowledge tab from the main navigation.
  2. Select the card for the Knowledge you created. This opens its management view.
  3. In the left panel, select the + (Add) icon.
  4. A new chat panel opens, ready for queries.

Add files to the Knowledge​

A new Knowledge is empty. To fill it, connect a data source and sync its content. This is covered in its own guide: Integrate external data.

Result​

You have a configured Knowledge with AI Search enabled, a default Touchpoint, and an open chat. Once you add files, the AI can answer using your own data.