Short definitions of the terms you encounter in the LiKE interface — especially in Analytics and when working with AI.

#AI and search

Term What it means
Indexing (embeddings) Converting the article's text into numeric vectors that power semantic search. Done automatically when an article is created or updated. Without indexing, an article is not found by AI search.
Transcription Converting an audio recording into text. In the editor you can upload an audio note — LiKE turns it into text that is then used like a regular article.
AI search (RAG) A pipeline of 15 stages: query validation → preparation → index search → ranking → generating an answer with links to sources.
Quality Score An article quality rating of 0–100 across five components: completeness, structure, readability, coherence and usefulness. Helps you understand which articles are worth improving.
Gaps (Gap Alerts) User questions that the knowledge base has no answer for. A signal that you need to create or expand an article.

#Answer types

In Analytics → By result type each AI answer is tagged with one of the following types:

Type What it means
AI-generated The answer was created from scratch by the model based on the sources found.
From cache A previously saved answer to a similar question is reused — without a new call to the model. Saves tokens and speeds up the answer.
Technical answer The model did not give a regular answer — for example, there is no relevant information in the base or the AI is temporarily unavailable. Direct excerpts from the found articles are shown.

#Tokens and costs

A token is a unit of text processing by an AI model (roughly ¾ of a word). By cost type:

Cost type What it means
Input tokens The text of the query and context sent to the model.
Output tokens The text of the answer generated by the model.
Indexing tokens Spent building vectors when indexing articles (see "Indexing").
Cache Hit Rate The share of answers served from cache. The higher it is, the lower the token consumption.

#Knowledge base structure

Term What it means
Space An isolated area of knowledge with its own articles, members and permissions. Search always stays within a single space.
Collection A folder for grouping articles within a space. Supports nesting and drag-and-drop.
Visibility (publish state) The access level of an article: Public (visible to everyone, including anonymously), Internal (only the team — the default), Draft (only editors) or Inherited (takes the level from the parent collection).
Tip: if you come across an unfamiliar term in the interface — find it here via AI search over this base.