Welcome to Intric Support Center
Introduction
Glossary and terms

Glossary and terms

Important terms you’ll encounter in Intric:

Intric-specific terms

  • Assistant: A customized AI chatbot that you configure for a specific task. It consists of a language model, a prompt, and connected knowledge.
  • Space: A digital collaboration area (think “folder” or “team”) where you and your colleagues gather your assistants and your shared knowledge.
  • Knowledge: Your internal information (documents, web pages, integrations) that’s uploaded to a Space. The assistant uses this as a library to find facts.
  • Collection: A folder within Knowledge where you organize files that belong together.
  • Security classification: Within Intric, information is divided into different security levels. This controls which AI models can be used and where data can be processed, so you never accidentally send sensitive information to the wrong place.

AI terms

  • Language model (LLM): The “brain” or engine of AI (e.g., GPT-4 or Claude). It’s trained on enormous amounts of text to understand and generate language.
  • Prompt: The instruction or “job description” you give to the assistant. It controls how the assistant should behave and respond.
  • RAG (Retrieval-Augmented Generation): The technique Intric uses to let the assistant read your data. The AI first searches for the right information in your Knowledge and then uses it to formulate the answer, instead of just guessing.
  • Context window (Working memory): The context window is the assistant’s working memory—how much information it can “keep in mind” simultaneously during an ongoing chat. If you send too much information at once, the assistant can start to “lose track.”
  • Tokens: How AI models count text. A word is often divided into multiple tokens. It’s used to measure how much text the model can handle at a time.
  • Hallucinations: Since AI works statistically, it can sometimes make up facts that sound very credible. This happens most often when it doesn’t have access to the right sources but still tries to answer your question. Using Knowledge in Intric significantly reduces the risk of this.