Azure AI Content Safety SDK for JavaScript

Introduction to Azure AI Content Safety SDK

The Azure AI Content Safety SDK for JavaScript empowers developers to build applications that can detect and mitigate harmful content. Leverage powerful AI models to analyze text and images for various categories of risks, ensuring a safer online environment for your users.

This SDK integrates seamlessly with Azure Content Safety, a cloud-based service designed to identify and flag potentially offensive, hateful, or unsafe content. Whether you're moderating user-generated content, ensuring brand safety, or complying with regulatory requirements, this SDK provides the tools you need.

Key Features:

• Text Moderation: Analyze text for hate speech, violence, sexual content, and self-harm.
• Image Moderation: Detect unsafe or inappropriate content within images.
• Customizable Severity Levels: Understand the risk associated with flagged content.
• Asynchronous Operations: Efficiently process large volumes of data.
• Easy Integration: Simple API design for quick adoption.

Installation

Install the Azure AI Content Safety SDK for JavaScript using npm or yarn:

npm install @azure/cognitiveservices-content-safety

yarn add @azure/cognitiveservices-content-safety

Getting Started

To start using the Content Safety client, you'll need an Azure subscription and a Content Safety resource. You can create these resources through the Azure portal.

Here's a basic example of how to initialize the client and analyze text:

import { TextAnalysisClient, AnalyzeTextRequest, SeverityLevel } from "@azure/cognitiveservices-content-safety";
import { TextAnalysisFromText } from "@azure/cognitiveservices-content-safety/models";

async function analyzeText(endpoint, apiKey, textToAnalyze) {
    const client = new TextAnalysisClient(endpoint, apiKey);

    const textAnalyzeRequest: AnalyzeTextRequest = {
        text: textToAnalyze,
        categories: ["Hate", "SelfHarm", "Sexual", "Violence"],
        severityLevel: SeverityLevel.High
    };

    const result = await client.analyzeText(textAnalyzeRequest);

    console.log(JSON.stringify(result, null, 2));
}

// Replace with your actual endpoint and API key
const CONTENT_SAFETY_ENDPOINT = "YOUR_CONTENT_SAFETY_ENDPOINT";
const CONTENT_SAFETY_API_KEY = "YOUR_CONTENT_SAFETY_API_KEY";
const MY_TEXT = "This is a sample text that may contain harmful content.";

analyzeText(CONTENT_SAFETY_ENDPOINT, CONTENT_SAFETY_API_KEY, MY_TEXT);
            

Core Concepts

• Categories: The types of harmful content you want to detect (e.g., Hate, Sexual, Violence, SelfHarm).
• Severity Level: A threshold to determine the level of severity for flagged content.
• Text Analysis: Detecting harmful content in textual data.
• Image Analysis: Detecting harmful content in images.

API Reference

class TextAnalysisClient {
    constructor(endpoint: string, apiKey: string);
    analyzeText(request: AnalyzeTextRequest): Promise<AnalyzeTextResponse>;
    analyzeImage(request: AnalyzeImageRequest): Promise<AnalyzeImageResponse>;
}
                

interface AnalyzeTextRequest {
    text: string;
    categories?: ("Hate" | "SelfHarm" | "Sexual" | "Violence")[];
    severityLevel?: SeverityLevel;
}
                

interface AnalyzeImageRequest {
    content: string; // Base64 encoded image
    content-type: "image/jpeg" | "image/png" | "image/gif" | "image/bmp";
    categories?: ("Hate" | "SelfHarm" | "Sexual" | "Violence")[];
    severityLevel?: SeverityLevel;
}
                

enum SeverityLevel {
    VeryLow,
    Low,
    Medium,
    High
}
                

Examples

Analyzing Image Content

To analyze an image, you need to provide the image data as a base64 encoded string.

import { ImageAnalysisClient, AnalyzeImageRequest, SeverityLevel } from "@azure/cognitivesservices-content-safety";
import * as fs from 'fs';

async function analyzeImage(endpoint, apiKey, imagePath) {
    const client = new ImageAnalysisClient(endpoint, apiKey);

    const imageBuffer = fs.readFileSync(imagePath);
    const base64Image = imageBuffer.toString('base64');

    const imageAnalyzeRequest: AnalyzeImageRequest = {
        content: base64Image,
        'content-type': 'image/jpeg', // or 'image/png', etc.
        categories: ["Sexual", "Violence"],
        severityLevel: SeverityLevel.Medium
    };

    const result = await client.analyzeImage(imageAnalyzeRequest);

    console.log(JSON.stringify(result, null, 2));
}

// Replace with your actual endpoint, API key, and image path
const IMAGE_ENDPOINT = "YOUR_CONTENT_SAFETY_ENDPOINT";
const IMAGE_API_KEY = "YOUR_CONTENT_SAFETY_API_KEY";
const MY_IMAGE_PATH = "./path/to/your/image.jpg";

// analyzeImage(IMAGE_ENDPOINT, IMAGE_API_KEY, MY_IMAGE_PATH);
            

Advanced Usage

Explore more advanced features such as handling custom categories, configuring retry policies, and integrating with other Azure services. For detailed information and best practices, please refer to the official Azure Content Safety documentation.