# Knowledge base

> The maintained, structured store of facts a system answers from — documents, prices, policies — that a chatbot grounds on so it can only say what is true.

Every grounded AI feature is only as good as its knowledge base: if the price list is stale, the model quotes stale prices fluently. The work is unglamorous — deciding which documents are authoritative, keeping them current, chunking them so the right fragment can be found — and it is most of the project.

A good one carries metadata as well as text: which product, which language, valid from when. Those fields are what let a system answer 'what does it cost' with this year's price instead of a confident average of every price ever written down.

## Related terms

- https://dfieldsolutions.com/en/glossary/rag.md
- https://dfieldsolutions.com/en/glossary/ai-chatbot.md
- https://dfieldsolutions.com/en/glossary/vector-database.md

---

Source: https://dfieldsolutions.com/en/glossary/knowledge-base
DField Solutions — Dunakeszi, Hungary — dezso@dfieldsolutions.com
Booking: see https://dfieldsolutions.com/en/contact