Knowledge
In the context of AI, knowledge is the organised, domain‑specific information a model can draw on to answer questions, separate from whatever it absorbed during training. It is the difference between what a model knows in general and what your organisation knows in particular.
A language model's built‑in knowledge is broad but fixed, frozen at the moment its training ended, and unaware of your products, policies or records. A knowledge base fills that gap with current, proprietary information the model can consult at answer time.
Built‑in knowledge vs a knowledge base
| Model's built‑in knowledge | A knowledge base | |
|---|---|---|
| Source | General training data | Your own documents and data |
| Freshness | Fixed at training time | Updated whenever you add to it |
| Scope | Broad and generic | Specific to your domain |
| Traceability | Hard to attribute | Answers can cite their source |
This is why grounding answers in a knowledge base, the core idea behind RAG, produces responses that are more accurate, more current and easier to trust.
How knowledge becomes usable
Raw documents are not directly searchable by a model. To make them useful, content is split into chunks, turned into embeddings and enriched with annotations, then stored in a vector database so the most relevant pieces can be retrieved on demand. Relationships between resources can also be captured, which is what enables graph‑based retrieval.
Knowledge in Rational AI
The Knowledge module is where you build and maintain this information: you ingest resources, organise them with tags and categories, view how they relate in a graph, and prepare them for retrieval. Kept up to date through sources, it becomes the foundation your agents reason over.
Additional resources
- Knowledge module: build and maintain your organisation's knowledge.
- Embeddings, chunks and annotations: how documents are prepared for retrieval.
- Retrieval-Augmented Generation: how knowledge is brought into an answer.