> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pixelpatrol.net/llms.txt
> Use this file to discover all available pages before exploring further.

# Moderation

> How content moderation works in Pixel Patrol

## Overview

Moderation is the core process of analyzing and filtering user-generated content. Pixel Patrol combines AI-powered analysis with rule-based decision making to provide flexible, accurate content moderation.

## Moderation Flow

```mermaid theme={null}
graph LR
    A[Media Submission] --> B[AI Analysis]
    B --> C[Rule Evaluation]
    C --> D{Decision}
    D -->|Approved| E[Content Allowed]
    D -->|Rejected| F[Content Blocked]
    D -->|Flagged| G[Manual Review]
    G --> H[Human Decision]
```

## Submission Process

### Media Types Supported

<CardGroup cols={3}>
  <Card title="Images" icon="image">
    * JPEG, PNG, GIF
    * WebP, AVIF
    * Up to 50MB
  </Card>

  <Card title="Videos" icon="video">
    * MP4, WebM
    * MOV, AVI
    * Up to 200MB
  </Card>

  <Card title="Text" icon="text">
    * Plain text
    * Rich text
    * Up to 1MB
  </Card>
</CardGroup>

### Submission Methods

1. **API Submission**: Direct API calls with media URLs
2. **File Upload**: Direct file upload to storage
3. **Batch Processing**: Multiple items in one request

## AI Analysis

### Detection Categories

Pixel Patrol's AI analyzes content for:

* **Violence**: Graphic violence, gore, weapons
* **Adult Content**: Nudity, sexual content
* **Hate Speech**: Discriminatory content, symbols
* **Self-Harm**: Content promoting self-injury
* **Drugs**: Drug use or paraphernalia
* **Spam**: Repetitive or promotional content
* **Custom Labels**: Team-specific categories

### Confidence Scores

Each detection includes a confidence score (0-1):

* `0.0-0.3`: Low confidence
* `0.3-0.7`: Medium confidence
* `0.7-1.0`: High confidence

## Moderation Queue

### Queue Management

The moderation queue shows:

* Pending items awaiting review
* AI analysis results
* Applied rules and decisions
* Moderation history

### Filtering Options

* **Status**: Pending, Approved, Rejected, Flagged
* **Content Type**: Image, Video, Text
* **Date Range**: Custom time periods
* **Confidence**: Score thresholds
* **Labels**: Specific AI detections

## Manual Review

### When Manual Review Occurs

* Items flagged by rules
* Low AI confidence scores
* Content in gray areas
* Random sampling for quality

### Review Actions

Moderators can:

* **Approve**: Allow the content
* **Reject**: Block the content
* **Escalate**: Send to senior moderator
* **Re-moderate**: Trigger new AI analysis

## Moderation Results

### Result Structure

```json theme={null}
{
  "id": "media_123",
  "status": "rejected",
  "ai_results": {
    "labels": [
      {
        "name": "violence",
        "confidence": 0.89
      }
    ]
  },
  "applied_rules": [
    {
      "rule_id": "rule_456",
      "rule_name": "Block Violence",
      "action": "reject"
    }
  ],
  "moderated_at": "2024-01-20T10:30:00Z",
  "moderator": "system"
}
```

### Webhook Notifications

Results are sent to configured webhooks:

```json theme={null}
{
  "event": "media.moderated",
  "media_id": "media_123",
  "status": "rejected",
  "timestamp": "2024-01-20T10:30:00Z"
}
```

## Performance Metrics

### Key Metrics

* **Processing Time**: Average time from submission to decision
* **Accuracy Rate**: Correct decisions vs. manual overrides
* **False Positive Rate**: Safe content incorrectly blocked
* **False Negative Rate**: Harmful content incorrectly approved

### Monitoring Dashboard

Track moderation performance:

* Real-time processing stats
* Decision distribution
* Rule effectiveness
* AI accuracy metrics

## Best Practices

### Optimization Tips

1. **Rule Tuning**: Adjust thresholds based on results
2. **Regular Reviews**: Sample and verify decisions
3. **Feedback Loop**: Learn from manual overrides
4. **Balance Speed/Accuracy**: Configure for your needs

### Common Patterns

1. **Strict First**: Start with strict rules, then relax
2. **Category Focus**: Prioritize high-risk content types
3. **User Trust**: Consider user reputation in rules
4. **Time-Based**: Adjust rules for peak periods

## Related Concepts

* [AI Moderation](/features/ai-moderation) - AI capabilities in detail
* [Rules](/concepts/rules) - Rule configuration
* [Queue Management](/tutorials/media-moderation/review-queue) - Using the moderation queue