Glossary and terms

The terms below carry a specific meaning in Intric. Where a word is also a label in the interface, this glossary uses it exactly as the platform shows it.

Product terms

• Assistant — a configurable AI colleague made of a language model, instructions, and knowledge. Can be personal (one per user, configured centrally) or shared inside a Space.
• Space — a shared work area where assistants live, members collaborate, and a security classification applies.
• Instructions — the baseline behavior you give an assistant: how it should act, what it’s for, and how it stays grounded. Called Prompt in earlier versions of Intric.
• Skill — a named procedure an assistant can recognize and follow, in addition to its general instructions.
• Knowledge — anything an assistant can draw on beyond the model’s own training: attachments, knowledge sources, integrations, custom tools, or web search. See Knowledge.
• Attachment — a file loaded onto an assistant permanently, available in every conversation with it.
• Knowledge source — material an assistant searches on demand rather than keeping loaded at all times.
• Conversation — a single exchange between you and an assistant. Collected under Conversations.
• Artefact — a document, presentation, spreadsheet, or other file an assistant produces, distinct from an ordinary chat answer. See Artefacts.
• Plan — a multi-step workflow an assistant runs in the background, on demand or on a schedule. See Plans.
• Widget — a chat window embedding one assistant on a public website. See Widgets.
• Integration — a connection to a service your organization already uses, such as SharePoint or Google Workspace.
• Custom tool (MCP) — a connection to your organization’s own systems, built using the Model Context Protocol.
• Security classification — a level assigned to a Space that determines which models and tools may be used in it, based on how sensitive the data is.
• Role — what a person can do. Organization roles are User, Creator and Admin; Space roles are Space Viewer, Space Editor and Space Admin. See Roles in Intric.
• Billing group — a cost center that groups users (and widgets) so their AI consumption can be tracked and allocated together.

AI terms

• Language model (LLM) — the engine behind an assistant. It’s trained on large amounts of text to understand and generate language.
• Retrieval-Augmented Generation (RAG) — the underlying technique that lets an assistant search your knowledge and use what it finds to ground its answer, instead of relying only on what the model already knows. See How Intric searches your documents.
• Context window — how much information a model can hold in mind at once during a conversation. Send too much at once and a model can start to lose earlier context.
• Tokens — the units models use to measure and price text. A word is often split into more than one token.
• Hallucination — a confident-sounding answer that isn’t actually correct. It happens most often when a model has no relevant knowledge to draw on but answers anyway. Attaching the right knowledge substantially reduces the risk.

Legacy terms

• Collection — a folder of documents inside the older, space-level Knowledge area. Being retired in favor of attachments. See Legacy knowledge.
• Website (crawled) — a website indexed for search under the older Knowledge area, part of the same legacy system as Collections.