Knowledge Health
Knowledge Health shows where the graph may need attention: possible duplicates, superseded material, contradictions, stale derivations, and analysis that failed or remains incomplete.
Open Knowledge → Health for the organization summary and scoped exception groups.
What it checks
| Signal | Why it matters |
|---|---|
| Possible duplicate | Two records may represent the same entity or fact. |
| Supersession | Newer knowledge may replace an older conclusion or artifact. |
| Contradiction | Supported statements disagree and should remain visible for review. |
| Stale derivation | A conclusion depends on a premise that changed or was invalidated. |
| Incomplete or failed analysis | A sweep did not finish or could not classify the available evidence. |
Summary states
The summary distinguishes a clean result from missing or failed evidence. A zero count is only healthy when the underlying analysis completed.
never_run— no health sweep has completed.healthy_shadow— analysis completed in recommendation-only mode with no actionable exceptions.healthy_active— active-mode evidence completed with no actionable exceptions.incomplete— some required analysis has not completed.action_needed— one or more exceptions need review.degraded— failures prevent the system from making a healthy claim.
Scope and evidence
Exception groups can be filtered by status, kind, Project, or Space. Results carry the evidence needed to understand why an item was surfaced. Knowledge Health is an organization dashboard surface and requires a signed-in user JWT; API keys cannot call these endpoints.
Availability
Knowledge Health is available in the dashboard and through its read-only JWT Management API. The endpoints require a signed-in user JWT and reject API keys. See the API Reference.