InworldSafetyModule | Inherits from: InworldFrameworkModule
Module for content safety checking and moderation within the Inworld framework. Provides real-time content analysis to detect potentially harmful or inappropriate content. Integrates with text embedding services for advanced semantic analysis of user inputs and AI responses.
Properties
Methods
Reference
SafetyConfig
Gets the current safety configuration instance. Contains the configured safety thresholds and topic settings for content moderation.Returns
Type:SafetyConfig
Description: The safety configuration instance with current threshold settings.
CreationConfig
Gets the safety checker creation configuration instance. Contains settings used during the initialization of the safety checker.Returns
Type:SafetyCheckerCreationConfig
Description: The creation configuration instance for safety checker initialization.
SetupSafetyThreshold
Sets up safety thresholds for content moderation. Replaces current threshold configuration with the provided data.Parameters
| Parameter | Type | Description |
|---|---|---|
| safetyData | List<SafetyThreshold> | List of safety thresholds to configure for different topic types. |
Returns
Type:void
SetupEmbedder
Sets up the text embedder interface for semantic analysis. Required for advanced content understanding and safety classification.Parameters
| Parameter | Type | Description |
|---|---|---|
| textEmbedderInterface | TextEmbedderInterface | The text embedder interface to use for content analysis. |
Returns
Type:void
SetupSafetyConfig
Configures safety settings based on provided threshold data. Converts SafetyThreshold objects to internal TopicThreshold format.Parameters
| Parameter | Type | Description |
|---|---|---|
| safetyThresholds | List<SafetyThreshold> | List of safety thresholds to configure. |
Returns
Type:void
IsSafe
Checks if the provided sentence is safe according to current safety thresholds. Performs real-time content analysis using the configured safety parameters.Parameters
| Parameter | Type | Description |
|---|---|---|
| sentence | string | The text content to analyze for safety violations. |
Returns
Type:bool
Description: True if the content is considered safe, false if it violates safety thresholds.
Serialized Fields
The following fields are configurable in the Unity Inspector:Safety Configuration
- m_SafetyData (
List<SafetyThreshold>) - List of safety thresholds for different topic types
Remote Configuration
- m_Provider (
string) - The AI model provider for safety checking - m_ModelName (
string) - The specific model name to use for safety analysis
Local Configuration
- m_ModelPath (
string) - The file path to a local safety model
Safety Threshold Configuration
TheSafetyThreshold class defines safety parameters for content moderation:
SafetyThreshold Properties
- topic (
UnsafeTopic) - The type of unsafe topic to monitor - threshold (
float) - The threshold value for this topic (typically between 0.0 and 1.0)
Threshold Behavior
- Lower values (closer to 0.0) are more restrictive and will flag more content as unsafe
- Higher values (closer to 1.0) are more permissive and will allow more content through