Skip to main content
Dome allows you to create custom detectors to implement domain-specific detection logic. This guide covers creating, registering, and using custom detectors.

​
Detection Categories

Register your detector with one of these categories:
CategoryUse Case
DetectionCategory.SecurityAdversarial attacks, injections
DetectionCategory.ModerationHarmful content, toxicity
DetectionCategory.PrivacyPII, secrets, sensitive data
DetectionCategory.IntegrityData quality, format validation
DetectionCategory.GenericAnything else
The category determines which guard types can use your detector.

​
Detection Result Format

The detect method must return a tuple of (flagged: bool, metadata: dict)
Always include query_string and response_string in metadata for proper guardrail operation.

​
Using Custom Detectors

  • Define before instantiation: custom detectors must be defined before creating the Dome instance.
  • Import from separate file: for cleaner organization, define detectors in a separate module.
Import your custom detector module before creating the Dome instance. The @register_method decorator must run before the config is parsed.

​
Example Detectors

  • Rate Limiter: Track request frequency per user
  • Regex Pattern Detector: Block content matching patterns
  • External API Detector: Call an external service for detection
  • Semantic Similarity Detector: Block content similar to known bad examples

Work in Progress

The programmatic protection capabilities and Dome integrations are currently in private preview and subject to change.

​
Next Steps

Configure Guardrails

Use custom detectors in configurations

Use Guardrails

Runtime integration patterns

Observability

Monitor custom detector performance

Protection Overview

Built-in detectors reference
Last modified on April 20, 2026