Core Concepts
Tukun.ai keeps the object model intentionally small. If users cannot explain what they are creating and reusing, the product becomes harder to govern.
The concepts that matter most
Section titled “The concepts that matter most”Account
Section titled “Account”An account is the main boundary for ownership in Tukun.ai.
An account owns:
- data sources
- conversations
- semantic assets
- cards
- dashboards
- billing and usage
If something should not be mixed together operationally, it probably belongs in a different account.
Workbench
Section titled “Workbench”The Workbench is the only execution surface. It is where you ask questions, inspect results, and decide what is trustworthy enough to save.
This is important because Tukun.ai does not split execution across multiple disconnected analysis surfaces.
Data source
Section titled “Data source”A data source is the approved connection to business data used for analysis. Questions run against one selected source at a time.
That means source choice is part of the answer context, not a background implementation detail.
Conversation
Section titled “Conversation”A conversation captures the sequence of a question, its follow-ups, and the resulting analytical context.
Use conversations for:
- exploration
- iteration
- narrowing ambiguity
- preserving context before a result is promoted into a reusable asset
Semantic asset
Section titled “Semantic asset”A semantic asset is a reusable definition of business meaning, such as a metric, dimension, or relationship that should stay consistent across repeated questions.
This is what turns “what did the model guess I meant?” into “what does the team actually mean by this metric?”
A card is a saved analytical output. Use it when a result is stable enough to be referenced again outside the original conversation.
Dashboard
Section titled “Dashboard”A dashboard is a grouped operating view built from saved outputs. Dashboards are for repeated review, not for replacing the exploration workflow.
How the concepts fit together
Section titled “How the concepts fit together”The normal progression is:
- Connect a data source.
- Ask a question in the Workbench.
- Review the result in the conversation.
- Resolve ambiguity through semantic definitions.
- Save stable outputs as cards.
- Group repeated cards into dashboards.
Common misunderstanding to avoid
Section titled “Common misunderstanding to avoid”Do not think of Tukun.ai as “chat first, assets later if we remember.” The product is designed for the opposite direction: exploratory work should become reusable assets when it proves stable.