Core Concepts

Kyb models your organisation as a semantic graph of typed entities and relationships. This page explains the building blocks.

Entities

An entity represents a real-world thing in your architecture: a system, service, application, database, business process, person, team, or any other element worth tracking. Every entity has a name, an optional description, and a type.

Entity types follow the ArchiMate 3.2 standard. Common types include Application Component, Application Service, Technology Service, Business Process, and Business Actor. Kyb extends ArchiMate with specializations like Person, Team, Database, GKE Cluster, and Cloud Run Service for more precise classification.

ArchiMate Layers

ArchiMate organises architecture into three layers. Every entity belongs to exactly one:

Relationships

Relationships are typed connections between entities. They define how things depend on, compose, serve, or communicate with each other.

Kyb supports two categories of relationship:

Confidence and Review Status

Every entity carries a confidence score from 0 to 100% indicating how certain the system is about its classification. Entities created by AI inference or cloud discovery start with lower confidence and are marked as Unreviewed.

The Review Queue lets you systematically work through unreviewed entities, approving or correcting each one. Reviewed entities are marked with a checkmark on the graph canvas.

Provenance

Every fact in the graph carries provenance: how it got there. This lets you trace any entity or relationship back to its origin. Source types include:

• -- LLM Inference (created by the AI assistant)
• -- GCP Discovery (auto-discovered from cloud)
• -- Archi Import (imported from an Archi file)
• -- Manual entry (created by a user)

Projects and Sub-projects

A project is a workspace containing a complete architecture graph. Projects can be organised hierarchically using sub-projects to represent bounded contexts or organisational divisions.

Entities from one project can reference entities in another project. These cross-project references appear as foreign nodes on the graph canvas, visually distinguished so you can see what belongs to the current project versus what is referenced from elsewhere.