Documentation

Public API Documentation

Public integration guides, API reference, and project setup notes for teams evaluating or implementing Paxmod. Learn about API performance optimization.

Quickstart Projects API Reference

Quickstart

Start moderating your projects in under five minutes.

1. Create a Project

Essential

The first step to using Paxmod is to create a project. Use the Projects tab on the left, open the Add New menu, and choose a moderation type (Text or Image). Name your project and add a description.

The project description is for your reference only and is not used by the moderation service to analyze or moderate content.

After creating your project, you'll need to configure your moderation settings. Text projects also choose a mode —

standard

advanced

Pax supports both text and image moderation. Image projects use category-based moderation (no advanced policy mode).

2. Adjust Your Project Settings

In standard mode, you can set the thresholds for toxicity, profanity, hate speech, and more. You likely will need to iterate on the thresholds to get the desired results.

The following categories can be configured:

Sexual
Hate and Discrimination
Violence
Dangerous and Criminal Content
Self-Harm
Medical Advice
Financial Advice
Legal Advice
Personally Identifiable Information
Harassment
Threatening
Profanities
Spam

3. Submit Content for Moderation

Integrate Pax into your application using our simple API. Send content for moderation before it's published to ensure a safe environment for your users.

Example Integration

// Before publishing user content
const checkContent = async (content) => {
  const response = await fetch('https://www.paxmod.com/api/v1/text', {
    method: 'POST',
    headers: {
      'Authorization': 'Bearer YOUR_API_KEY',
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      message: content,
      user_id: 'optional_user_id',
      context_id: 'optional_context_id'
    })
  });

  const result = await response.json();
  
  if (result.content_moderation?.flagged) {
    // Handle flagged content
    return false;
  }
  
  // Content is safe to publish
  return true;
}

4. Review Flagged Messages

Monitor and review flagged messages in your analytics dashboard. You can:

View messages flagged by category
Track messages by user ID and context ID
Provide feedback on moderation accuracy
Track moderation performance over time

All Done!

You're done! You are now ready to moderate content using Pax. If you have any feedback or questions, please shoot us an email at contact@paxmod.com

Was this page helpful?

Projects

Learn how to create and manage your content moderation projects.

Text Projects

Set up text content moderation projects.

Image Projects

Set up AI-powered image content moderation.

Project Configuration

Projects can be configured in two modes: standard or advanced. Each mode offers different levels of control over content moderation.

Standard Mode

standard

Configure individual thresholds for specific content categories. Ideal for most use cases where you need granular control over different types of content.

Advanced Mode

advanced

Define custom moderation policies using natural language. Perfect for complex use cases requiring specific content guidelines.

Features:

Custom policy definitions
Single threshold setting
Natural language rules
Complex policy enforcement

Project Features

Threshold Management

Adjust sensitivity thresholds (0-1) for each category or policy. Lower values are more permissive, higher values are more strict. The playground allows real-time testing of threshold adjustments.

Whitelist System

Add terms or phrases that should never be flagged. Useful for game-specific terminology, character names, or approved content that might otherwise trigger moderation.

Project Playground

Test your moderation settings in real-time. View detailed category scores, adjust thresholds, and see immediate results. Changes in the playground can be saved using the Update Project button.

Image Moderation

Image moderation uses our advanced multimodal AI models to detect harmful visual content. Unlike text moderation, image projects only support category-based moderation (no advanced mode or whitelist).

Image moderation supports a dedicated set of visual categories (including Sexual/Minors, Violence/Graphic, Self-Harm/Intent/Instructions, Harassment, Hate, and Illicit). Choose the categories you want to enforce when creating the project.

Image Categories

Image moderation analyzes visual content across the following categories:

Sexual / Sexual Minors

Detects sexually explicit or suggestive imagery, including adult content and inappropriate poses.

Violence / Violence Graphic

Identifies violent imagery including weapons, physical harm, gore, and graphic violence.

Self-Harm / Intent / Instructions

Detects depictions of self-harm, including cutting, suicide-related imagery, and eating disorders.

Harassment / Hate / Illicit

Flags hateful symbols, harassment/threats, and illicit activity when present in images.

Creating an Image Project

Navigate to the Projects section in your console
Open the "Add New" menu and choose "Image moderation"
Choose which image categories to enable
Save and get your API key
Fine-tune category thresholds later from the project's page (each defaults to 0.5)

Important Limitations

• No Advanced Mode: Image projects only support category-based moderation
• No Whitelist: URL-based whitelisting isn't supported for images
• Different Categories: Image categories differ from text (e.g., Sexual/Minors, Violence/Graphic, Illicit)

Fine-tuning Your Project

It's recommended to test and adjust your thresholds in the project playground before deploying to production.

Adjusting Category Thresholds

Each moderation category has a threshold slider ranging from 0 to 1:

0.0-0.3: Very permissive (allows most content)
0.4-0.6: Balanced moderation (recommended starting point)
0.7-1.0: Strict moderation (flags more content)

Managing Categories

In standard mode, you can enable or disable specific moderation categories based on your needs. Disabled categories will not be checked during content moderation.

Disabling categories can improve performance but may create gaps in your content moderation. Consider your use case carefully.

Allow List

The allow list tells Paxmod how to handle terms that are safe in your community but look risky to a general-purpose classifier. Useful for game mechanics, character names, and domain vocabulary.

Each term carries its own behaviour:

Skip moderation — the message is returned unflagged with no category scores at all. Every other category is skipped too.
Remove word — the term is dropped and the rest of the message is still checked normally.
Replace word — the term is swapped for a neutral word you choose, then the message is checked. Usually the best result, because the sentence still reads naturally to the classifier.

Matching is whole-word and case insensitive, and applies in both standard and advanced mode. When a term is removed or replaced, the response includes an allow_list field showing what was applied, and processed_message shows the exact text that was scored.

Skip moderation creates a blind spot: any message containing that term passes untouched, however else it reads. Prefer remove or replace unless you specifically want the whole message exempted.

Testing Your Configuration

Use the project playground to:

Test different types of content
Adjust thresholds in real-time
View detailed category scores
Verify allow list behaviour

It's recommended to test with a variety of content samples that represent your actual use case, including edge cases.

Content Flagging Logic

Messages are flagged when their category score exceeds the defined threshold. Default threshold is 0.5 if not specified.

In standard mode, content is analyzed across multiple categories. Each enabled category has its own threshold and scoring:

Each category receives a score between 0 and 1
Content is flagged if any category score exceeds its threshold
Profanity detection uses a rule-based filter that catches obfuscated profanities
Other categories use AI classifier scoring

Standard Mode Response Example

{
  "status": "success",
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "service": "pax",
  "result": "flagged",
  "content_moderation": {
    "flagged": true,
    "categories": {
      "Harassment": {
        "flagged": true,
        "score": 0.8234,
        "threshold": 0.5
      },
      "Profanities": {
        "flagged": true,
        "score": 0.8,
        "threshold": 0.5
      },
      "Threatening": {
        "flagged": false,
        "score": 0.2156,
        "threshold": 0.5
      }
    }
  }
}

Was this page helpful?

API Documentation

Complete API documentation for integrating Pax content moderation.

Authentication

Required

All API requests require authentication using your API key in the Authorization header (recommended). For backward compatibility, you may instead pass it as an api_key field in the request body.

Authentication Header

Authorization: Bearer YOUR_API_KEY

Keep your API key secure and never expose it in client-side code.

Text Moderation Endpoint

Send content for moderation using our text moderation endpoint. The API supports both standard and advanced modes based on your project configuration.

POST /api/v1/text

Request Example

const response = await fetch('https://www.paxmod.com/api/v1/text', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    message: 'Text to moderate',
    user_id: 'optional_user_id',
    context_id: 'optional_context_id',
    advanced_threshold: 0.5
  })
});

const result = await response.json();

Request Parameters

{
  "message": "string (required)",     // The text content to moderate (max 5000 characters)
  "user_id": "string (optional)",     // Identifier for the message author
  "context_id": "string (optional)",  // Conversation ID (e.g., room, channel, private chat). Same ID for everyone in it
  "advanced_threshold": "number (optional)" // Custom threshold (0-1) for advanced mode
}

Response Format

{
  "status": "success",
  "id": "unique-moderation-id",
  "service": "pax",
  "request": {
    "original_message": "Text that was moderated",
    "processed_message": "Normalized text",
    "timestamp": "2024-03-21T12:00:00.000Z"
  },
  "result": "flagged" | "not_flagged",
  "reason": "Reason for flagging or whitelisting",
  "degraded": false,             // true if a moderation provider was briefly unavailable
  "unavailable_categories": [],  // enabled categories that couldn't be scored while degraded
  "content_moderation": {
    "flagged": boolean,
    "categories": {
      "category_name": {
        "flagged": boolean,
        "score": number,
        "threshold": number
      }
    }
  }
}

Player Context

Player Context reads whole conversations to find risks that build up over many messages, like grooming. For it to work, send user_id (the player who wrote the message) and context_id (one ID per conversation) with every message. Use the same context_id for everyone in a conversation, including both players in a private chat. Otherwise only one side is visible.

Graceful degradation

Standard-mode moderation draws on several providers. If one is briefly unavailable, the request still succeeds (HTTP 200) instead of erroring. Any enabled category that couldn't be scored is omitted from content_moderation.categories and listed in unavailable_categories, and degraded is set to true. Those categories are not flagged (fail-open), so treat a degraded: true response as reduced coverage and apply extra caution if your use case requires it.

Error Responses

{
  "status": "error",
  "error": {
    "code": "ERROR_CODE",
    "message": "Human readable message",
    "details": "Detailed error description",
    "timestamp": "2024-03-21T12:00:00.000Z"
  }
}

HTTP	code	When
401	UNAUTHORIZED	No API key provided.
401	INVALID_API_KEY	API key is invalid or revoked.
400	EMPTY_MESSAGE	Message is empty after trimming.
400	MESSAGE_TOO_LONG	Message exceeds 5000 characters.
429	USAGE_LIMIT_EXCEEDED	Monthly message limit for your plan or trial reached.
503	LLM_SERVICE_ERROR	Advanced-mode language model was unavailable.
500	INTERNAL_SERVER_ERROR	Unexpected server error.

Image Moderation Endpoint

New

Moderate images by providing publicly accessible image URLs. The API uses our advanced vision models to detect sexual, violent, self-harm, harassment, hateful, and illicit visual content.

Image Moderation Requirements

Project must be configured with moderation_type: "image"
Categories include Sexual/Minors, Violence/Graphic, Self-Harm/Intent/Instructions, Harassment/Threatening, Hate/Threatening, and Illicit/Violent
Images must be publicly accessible via HTTPS URLs
Supports common formats: JPEG, PNG, GIF, WebP

POST /api/v1/image

Request Example

const response = await fetch('https://www.paxmod.com/api/v1/image', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    image_url: 'https://example.com/image.jpg',
    user_id: 'user123',          // optional
    context_id: 'upload_456'     // optional
  })
});

const data = await response.json();
console.log(data);

Request Parameters

Parameter	Type	Required	Description
image_url	string	Yes	Publicly accessible HTTPS URL of the image to moderate
user_id	string	No	Optional identifier for the user who uploaded the image
context_id	string	No	Optional context identifier (e.g., upload session, chat room)

Success Response

{
  "status": "success",
  "id": "12f8d080-021e-4a3d-8f07-bda632fca263",
  "service": "pax",
  "request": {
    "image_url": "https://example.com/image.jpg",
    "timestamp": "2025-10-21T20:12:38.507Z"
  },
  "result": "not_flagged",  // or "flagged"
  "reason": "No moderation category thresholds exceeded",
  "content_moderation": {
    "flagged": false,
    "categories": {
      "Sexual": {
        "flagged": false,
        "score": 0.0023,
        "threshold": 0.5
      },
      "Sexual/Minors": {
        "flagged": false,
        "score": 0.0001,
        "threshold": 0.5
      },
      "Violence/Graphic": {
        "flagged": false,
        "score": 0.0145,
        "threshold": 0.5
      },
      "Self-Harm/Intent": {
        "flagged": false,
        "score": 0.0012,
        "threshold": 0.5
      }
    }
  }
}

Error Responses

Invalid Project Type

{
  "status": "error",
  "error": {
    "code": "INVALID_MODERATION_TYPE",
    "message": "Project not configured for image moderation",
    "details": "This project is configured for text moderation. Please create a new project with image moderation type.",
    "timestamp": "2025-10-21T20:12:38.507Z"
  }
}

Missing or Invalid Image URL

{
  "status": "error",
  "error": {
    "code": "MISSING_IMAGE_URL",
    "message": "Image URL is required",
    "details": "Please provide an image_url in the request body",
    "timestamp": "2025-10-21T20:12:38.507Z"
  }
}

Was this page helpful?

1. Create a Project

2. Adjust Your Project Settings

3. Submit Content for Moderation

Example Integration

4. Review Flagged Messages

All Done!

Text Projects

Image Projects

Project Configuration

Standard Mode

Available Categories:

Advanced Mode

Features:

Project Features

Threshold Management

Available Categories:

Whitelist System

Project Playground

Image Moderation

Image Categories

Sexual / Sexual Minors

Violence / Violence Graphic

Self-Harm / Intent / Instructions

Harassment / Hate / Illicit

Creating an Image Project

Fine-tuning Your Project

Adjusting Category Thresholds

Managing Categories

Allow List

Testing Your Configuration

Content Flagging Logic

Standard Mode Response Example

Authentication

Authentication Header

Text Moderation Endpoint

POST /api/v1/text

Request Example

Request Parameters

Response Format

Error Responses

Image Moderation Endpoint

POST /api/v1/image

Request Example

Request Parameters

Success Response

Error Responses

Invalid Project Type

Missing or Invalid Image URL