glin-profanity/dist/types-Ct_ueYqw.d.cts

Glin-Profanity is a lightweight and efficient npm package designed to detect and filter profane language in text inputs across multiple languages. Whether you’re building a chat application, a comment section, or any platform where user-generated content

/**
 * Type definitions for ML-based profanity detection.
 */
/**
 * Toxicity categories detected by the TensorFlow.js model.
 * These map to the civil comments dataset labels.
 */
type ToxicityLabel = 'identity_attack' | 'insult' | 'obscene' | 'severe_toxicity' | 'sexual_explicit' | 'threat' | 'toxicity';
/**
 * Result from a single toxicity prediction.
 */
interface ToxicityPrediction {
    /** The toxicity category */
    label: ToxicityLabel;
    /** Whether the text matches this category (null if below threshold) */
    match: boolean | null;
    /** Probability scores [non-toxic, toxic] */
    probabilities: [number, number];
}
/**
 * Result from ML-based toxicity analysis.
 */
interface MLAnalysisResult {
    /** Whether any toxicity was detected */
    isToxic: boolean;
    /** Overall toxicity score (0-1) */
    overallScore: number;
    /** Predictions for each category */
    predictions: ToxicityPrediction[];
    /** Categories that matched */
    matchedCategories: ToxicityLabel[];
    /** Processing time in milliseconds */
    processingTimeMs: number;
}
/**
 * Configuration for the ML toxicity detector.
 */
interface MLDetectorConfig {
    /**
     * Minimum confidence threshold for predictions.
     * Values below this threshold will return null for match.
     * @default 0.85
     */
    threshold?: number;
    /**
     * Specific toxicity categories to check.
     * If not specified, all categories are checked.
     */
    labels?: ToxicityLabel[];
    /**
     * Whether to load the model immediately on instantiation.
     * If false, model will be loaded on first use.
     * @default false
     */
    preloadModel?: boolean;
}
/**
 * Combined result from both rule-based and ML detection.
 */
interface HybridAnalysisResult {
    /** Rule-based detection result */
    ruleBasedResult: {
        containsProfanity: boolean;
        profaneWords: string[];
    };
    /** ML-based detection result (null if ML not enabled) */
    mlResult: MLAnalysisResult | null;
    /** Combined decision */
    isToxic: boolean;
    /** Confidence score for the decision */
    confidence: number;
    /** Reason for the decision */
    reason: string;
}

export type { HybridAnalysisResult as H, MLDetectorConfig as M, ToxicityLabel as T, MLAnalysisResult as a, ToxicityPrediction as b };