Compliance Logging

Luna logs all clinical interactions and escalations for regulatory compliance and quality assurance.

Clinical Interaction Logging

Every conversation that involves clinical content is logged via logClinicalInteraction():


logClinicalInteraction({
  userId: user.id,
  sessionId: conversationId,
  interactionType: 'chat',
  toolsUsed: ['getLabResults', 'checkContraindication'],
  biomarkersDiscussed: ['testosterone-total', 'tsh'],
  peptidesDiscussed: ['bpc-157'],
  guardrailResult: {
    passed: true,
    violations: [],
  },
  timestamp: new Date().toISOString(),
});

What Gets Logged

Field	Description
`userId`	Clerk user ID
`sessionId`	Conversation session ID
`interactionType`	`chat`, `voice`, `support`
`toolsUsed`	Which Luna tools were invoked
`biomarkersDiscussed`	Biomarker codes mentioned
`peptidesDiscussed`	Peptides mentioned
`guardrailResult`	Post-response guardrail check outcome
`timestamp`	ISO timestamp

Escalation Logging

Escalations are logged when:

Emergency detection triggers
Guardrail violations are critical
Patient context contains emergency flags


logEscalation({
  userId: user.id,
  type: 'guardrail_violation',
  severity: 'critical',
  details: {
    layer: 'safety',
    violation: 'dosage_exceeded',
    message: 'Response suggested dosage above safety limit',
  },
  timestamp: new Date().toISOString(),
});

Escalation Types

Type	Trigger	Severity
`emergency_detected`	Crisis keywords in user message	Critical
`guardrail_violation`	Critical guardrail failure	Critical
`patient_emergency_flag`	Patient context has emergency flags	Critical
`emergency_dismissal`	User dismisses emergency concern	Warning

Guardrail-Triggered Logging

After every AI response, the guardrail result is checked:


const guardrailResult = await checkGuardrails(guardrailInput);
 
if (guardrailResult.shouldEscalate) {
  logEscalation({
    userId: user.id,
    type: 'guardrail_violation',
    severity: 'critical',
    details: guardrailResult.violations,
  });
}
 
logClinicalInteraction({
  ...interactionData,
  guardrailResult,
});

Compliance Reports

The compliance logging data supports:

Audit trail — Complete record of all clinical AI interactions
Quality review — Identify conversations with guardrail violations
Regulatory reporting — HIPAA-compliant interaction records
Safety monitoring — Track emergency detection and escalation patterns
Tool usage analytics — Understand which tools are most used

Data Flow


AI Response Generated
        │
        ▼
Guardrails Check ──→ Violations? ──→ logEscalation()
        │
        ▼
logClinicalInteraction()
        │
        ▼
Store in conversation_history
        │
        ▼
Record patient_event (if clinical)