# Coherence Profiles

A `CoherenceProfile` is a versioned desired state, not a description of an entity's inherent identity. It contains:

- subject entity and scope;
- purpose and lifecycle status;
- dimension configurations with target, tolerance, response zones, evidence override if permitted, and boundary references;
- principles, values, preferences, and deterministic constraints;
- policy references;
- Authority, Consent, visibility, retention, provenance, and supersession;
- activation and expiry.

## Dimension configuration

Each dimension configuration MUST name a `dimensionRef`, set a target point, set an inclusive tolerance interval, define ordered band-drift thresholds, and identify allowed responses per zone. A target point MUST fall inside its tolerance interval and the scale. Zones MUST be monotonic and MUST NOT cross outside scale distance.

Profiles contain desired targets only. Current estimates are stored in `CoherenceState`; comparisons are stored in `CoherenceAssessment`. This separation prevents an observation from silently rewriting an intention.

## Lifecycle

`draft → rehearsal → published → superseded | retired`

Drafts can be edited under configuration Permission. Rehearsal runs scenarios without live effects. Publication creates an immutable version and an auditable configuration Action. A published profile may be superseded, never rewritten.

## Human-related profiles

An Avatar profile uses first-person, contextual language: “how I want to show up here.” It MUST NOT be exported as the Source's personality profile, used to rank the Source, or treated as permission to change human behavior. The Source can reset observations and may retain the desired profile while deleting permitted derived history.