Skip to main content

​
AI Models Registry

The Models Registry is your centralized hub for managing all AI models used in your applications, including risk classification, compliance status, and performance monitoring.

​
Why Register Models?

EU AI Act Compliance

Required for high-risk AI systems under EU AI Act

Risk Classification

Automatically classify models by risk level

Documentation

Maintain technical documentation for each model

Monitoring

Track model performance and compliance over time

​
Risk Classifications

Models are classified according to EU AI Act risk levels:
  • Unacceptable Risk
  • High Risk
  • Limited Risk
  • Minimal Risk
Prohibited - Cannot be deployed
  • Social scoring systems
  • Subliminal manipulation
  • Exploitation of vulnerabilities
  • Real-time biometric identification (with exceptions)

​
Registering a Model

1

Basic Information

const model = await regpilot.models.register({
  name: "Customer Support Classifier",
  version: "1.0.0",
  provider: "openai",
  model_id: "gpt-4-turbo",
  description: "Classifies customer support tickets"
});
2

Risk Classification

Select the appropriate risk level based on intended use:
await regpilot.models.update(model.id, {
  risk_level: "limited",
  intended_purpose: "Customer support automation",
  target_users: "Customer service representatives"
});
3

Technical Documentation

Upload required documentation:
  • System architecture
  • Training data description
  • Performance metrics
  • Risk assessment
4

Compliance Settings

Configure compliance monitoring:
await regpilot.models.setCompliance(model.id, {
  frameworks: ["eu_ai_act", "gdpr"],
  monitoring_enabled: true,
  governor_enabled: true
});

​
Model Information

Each registered model includes:

​
Basic Details

  • Name and version
  • Provider and model ID
  • Description and purpose
  • Deployment environment

​
Risk Assessment

  • Risk classification
  • Intended purpose
  • Target user groups
  • Potential risks identified

​
Technical Documentation

  • Model architecture
  • Training data sources
  • Performance metrics
  • Validation results
  • Bias testing results

​
Compliance Status

  • Applicable frameworks
  • Compliance score
  • Active violations
  • Last assessment date

​
Managing Models

​
Update Model

await regpilot.models.update("model_123", {
  version: "1.1.0",
  status: "production",
  performance_metrics: {
    accuracy: 0.94,
    f1_score: 0.92,
    latency_p95: 250
  }
});

​
Archive Model

await regpilot.models.archive("model_123", {
  reason: "Replaced by v2.0",
  replacement_model_id: "model_456"
});

​
Generate Documentation

const docs = await regpilot.models.generateDocs("model_123", {
  format: "pdf",
  include_sections: [
    "technical_specs",
    "risk_assessment",
    "compliance_status",
    "performance_metrics"
  ]
});

​
Model Monitoring

Continuous monitoring includes:
  • Performance Tracking: Accuracy, latency, error rates
  • Compliance Validation: Ongoing EU AI Act checks
  • Bias Detection: Regular fairness assessments
  • Usage Analytics: Request volumes and patterns
  • Incident Tracking: Errors and violations

​
Compliance Requirements

For high-risk models, you must maintain:
  • Detailed system description
  • Development and testing procedures
  • Risk management documentation
  • Data governance records
  • Identified risks and mitigations
  • Residual risk evaluation
  • Testing and validation results
  • Ongoing monitoring procedures
  • Training data description
  • Data quality metrics
  • Bias assessments
  • Data provenance records
  • Oversight measures
  • Stop/override capabilities
  • Human-in-the-loop procedures
  • Responsibility assignment

​
Best Practices

1

Register Early

Register models during development, not just at deployment
2

Version Control

Track all model versions with detailed change logs
3

Regular Updates

Keep documentation and compliance status current
4

Monitor Continuously

Enable ongoing performance and compliance monitoring
5

Document Changes

Record all updates, incidents, and modifications

​
Next Steps

Risk Classification Guide

Learn about EU AI Act risk levels

Documentation Requirements

Technical documentation guide

API Reference

Models API documentation

Governor

Enable Governor validation