Implementation:Openai Openai node Moderations Resource

Knowledge Sources	Openai_Openai_node
Domains	SDK, Content Moderation, Safety
Last Updated	2026-02-15 12:00 GMT

Overview

The Moderations resource class provides a method for classifying text and image inputs as potentially harmful across multiple content categories.

Description

The Moderations class extends APIResource and exposes a single create method that sends content to the OpenAI moderation endpoint (/moderations). It accepts text strings, arrays of strings, or multimodal inputs combining text and images. The API returns detailed classification results across 13 content categories including harassment, hate speech, self-harm, sexual content, violence, and illicit activities.

Each moderation result includes three complementary data structures: categories with boolean flags indicating whether content was flagged in each category, category_scores with numerical confidence scores (0 to 1) for each category, and category_applied_input_types indicating which input types (text or image) contributed to each category score. The top-level flagged boolean provides a quick check for whether any category was triggered.

The module supports two model families: the omni-moderation models that handle both text and image inputs, and the text-moderation models for text-only classification. Image inputs are provided via the ModerationImageURLInput type, which accepts either direct URLs or base64-encoded data URLs.

Usage

Use this resource to screen user-generated content for policy violations before displaying or processing it. It is commonly integrated into content pipelines, chat applications, and user-facing platforms to detect harmful content such as harassment, hate speech, self-harm content, sexual material, violence, and illicit activity.

Code Reference

Source Location

• Repository: openai-node
• File: src/resources/moderations.ts

Signature

export class Moderations extends APIResource {
  create(body: ModerationCreateParams, options?: RequestOptions): APIPromise<ModerationCreateResponse>;
}

Import

import OpenAI from 'openai';

I/O Contract

Inputs

Name	Type	Required	Description
input	Array<string> | Array<ModerationMultiModalInput>	Yes	Text or multimodal content to classify
model	ModerationModel	No	Model to use: omni-moderation-latest, omni-moderation-2024-09-26, text-moderation-latest, or text-moderation-stable

Outputs

Name	Type	Description
id	string	Unique identifier for the moderation request
model	string	The model used to generate results
results	Array<Moderation>	List of moderation objects, one per input
results[].flagged	boolean	Whether any category is flagged
results[].categories	Moderation.Categories	Boolean flags for each of 13 content categories
results[].category_scores	Moderation.CategoryScores	Numerical scores (0-1) for each category
results[].category_applied_input_types	Moderation.CategoryAppliedInputTypes	Input types that each score applies to

Content Categories

Category	Description	Input Types
harassment	Harassing language towards any target	text
harassment/threatening	Harassment including violence or serious harm	text
hate	Hate based on protected characteristics	text
hate/threatening	Hateful content with violence towards targeted groups	text
illicit	Instructions for wrongdoing or illicit acts	text
illicit/violent	Illicit instructions including violence or weapons	text
self-harm	Content promoting acts of self-harm	text, image
self-harm/instructions	Instructions for self-harm acts	text, image
self-harm/intent	Expressions of self-harm intent	text, image
sexual	Sexually explicit content	text, image
sexual/minors	Sexual content involving minors	text
violence	Content depicting death, violence, or injury	text, image
violence/graphic	Graphic depictions of violence	text, image

Usage Examples

import OpenAI from 'openai';

const client = new OpenAI();

// Moderate a single text string
const result = await client.moderations.create({
  input: 'Some text to check for policy violations',
});

console.log(result.results[0].flagged);
console.log(result.results[0].category_scores.harassment);

// Moderate multiple text inputs
const batchResult = await client.moderations.create({
  input: ['First message', 'Second message'],
  model: 'omni-moderation-latest',
});

for (const moderation of batchResult.results) {
  if (moderation.flagged) {
    console.log('Content flagged:', moderation.categories);
  }
}

// Moderate multimodal input (text + image)
const multiModalResult = await client.moderations.create({
  input: [
    { type: 'text', text: 'Check this image' },
    { type: 'image_url', image_url: { url: 'https://example.com/image.png' } },
  ],
  model: 'omni-moderation-latest',
});

Key Types

ModerationModel

type ModerationModel =
  | 'omni-moderation-latest'
  | 'omni-moderation-2024-09-26'
  | 'text-moderation-latest'
  | 'text-moderation-stable';

ModerationMultiModalInput

type ModerationMultiModalInput = ModerationImageURLInput | ModerationTextInput;

interface ModerationImageURLInput {
  image_url: { url: string };
  type: 'image_url';
}

interface ModerationTextInput {
  text: string;
  type: 'text';
}

Related Pages

• Environment:Openai_Openai_node_Node_20_Runtime