Academy
2. Creating Your First Knowledge Base
Create a new knowledge base by choosing a name, description, embedding model, and understanding chunk size and overlap defaults.
A knowledge base is a private, indexed collection of your documents. When you create one, the platform splits your content into searchable pieces so the AI can retrieve relevant passages and ground its answers in your own material — whether you're working in ChatPro, OUPI Studio, Deep Research, or workflows. Think of it as giving the AI a personal library it can consult before answering you.
To create a knowledge base, click "New Knowledge Base". You'll need to provide:
• Name — a clear label so you (and teammates) can identify it later. Tip: keep one base per topic for sharper retrieval. • Description — a short summary of what the base contains; this helps you stay organised as your collection grows. • Embedding model — the model that converts your text into searchable vectors. Each base is tied to one embedding model chosen at creation. • Initial text content (optional) — you can paste text right away to seed the base.
You can also switch to MCP import mode at creation to pull content from a connected integration instead of uploading files, with automatic syncing.
Before indexing, every document is split into chunks. Two settings control this:
• Chunk size — how large each piece of text is. Larger chunks give the AI more context per result; smaller chunks are better for short, self-contained items like FAQ entries. • Overlap — how much text is shared between consecutive chunks. Overlap prevents ideas from being cut off at a boundary.
The defaults are designed to work well for most documents. Only adjust them if you have a specific document shape — for example, increase chunk size for long structured reports, or decrease it for collections of short answers.
Keep one knowledge base per topic rather than dumping everything into a single base. Focused bases produce sharper, more relevant retrieval results when the AI searches for answers.
Be aware that indexing consumes embedding credits proportional to the size of your content, and your administrator may have set a cap on the number of documents per base. Check your plan limits before bulk-uploading large document sets.
Open the Knowledge Bases page and click "New Knowledge Base". Give it a name and description, choose an embedding model, and leave chunk size and overlap at their defaults. Create the base — you'll use it to upload documents in the next lesson. Use the guided tour "Getting started with knowledge bases" if available.
You now know how to create a knowledge base: pick a clear name and description, choose an embedding model, and leave chunk size and overlap at their sensible defaults unless your documents have a special shape. You can optionally paste initial text or use MCP import mode. Each base is a private, indexed collection that the AI searches to ground its answers across the entire platform. Next, you'll learn how to add and manage documents inside your new base.
- Create a knowledge base — Start here: name it, pick an embedding model, and add your first content.
- Your knowledge at a glance — Bases, documents and indexing status are summarized here.
- Your knowledge bases — Open a base to add documents, test searches, and share it with your team.