Settings
System status and configuration
Connection Status
Backend API
Disconnected
Checking…
Phoenix Cloud
Disconnected
Checking…
Gemini API
Disconnected
Checking…
Database
Disconnected
Checking…
Configuration
Loading config…
Evaluator Registry (14 evaluators)
| Name | Type | Annotation Level |
|---|---|---|
| SchemaValidity | Code | span |
| ToolSequence | Code | session |
| RefundGuard | Code | span |
| PrivacyGuard | Code | span |
| EscalationGuard | Code | span |
| CitationPresence | Code | span |
| LatencyBudget | Code | span |
Judge calibration
Judge calibration
Cohen’s κ between each LLM judge and seeded canary labels. Landis & Koch: κ ≥ 0.6 = substantial agreement.
κ tones — brand ≥ 0.60 · ink 0.20–0.60 · fail < 0.20aggregate computed —
Diagnostic Actions
Phoenix Cloud workspace
Cross-reference everything PhoenixLoop logs by jumping to the live Phoenix workspace.