​
Overview

Rule-based moderation allows you to define custom logic for content filtering. Combined with AI analysis, rules provide precise control over moderation decisions based on your specific requirements.

​
Rule Engine

​
How Rules Work

The rule engine evaluates content through a series of conditions:

​
Rule Components

Each rule consists of:
ComponentDescriptionExample
ConditionsWhat to checkai.violence.confidence > 0.8
ActionsWhat to doreject, approve, flag
PriorityExecution order100 (lower = higher priority)
MetadataAdditional context{reason: "High violence"}

​
Rule Types

​
AI-Based Rules

Leverage AI analysis results: 
{
  "name": "Block High-Confidence Violence",
  "type": "ai_analysis",
  "conditions": {
    "ai.violence.confidence": { ">=": 0.8 }
  },
  "action": "reject"
}

​
Metadata Rules

Check content metadata: 
{
  "name": "Auto-Approve Verified Users",
  "type": "metadata",
  "conditions": {
    "user.verified": true,
    "user.trust_score": { ">=": 0.9 }
  },
  "action": "approve"
}

​
Composite Rules

Combine multiple conditions: 
{
  "name": "Complex Content Filter",
  "type": "composite",
  "conditions": {
    "OR": [
      {
        "AND": [
          {"ai.adult.confidence": { ">=": 0.7 }},
          {"metadata.context": { "!=": "educational" }}
        ]
      },
      {"ai.violence.confidence": { ">=": 0.9 }}
    ]
  },
  "action": "reject"
}

​
Time-Based Rules

Apply rules based on time: 
{
  "name": "Weekend Strict Moderation",
  "type": "temporal",
  "conditions": {
    "time.day_of_week": { "in": ["saturday", "sunday"] },
    "ai.any.confidence": { ">=": 0.5 }
  },
  "action": "flag"
}

​
Rule Groups

​
Organization

Rules are organized into groups for better management: 
interface RuleGroup {
  id: string;
  name: string;
  description: string;
  priority: number;
  active: boolean;
  rules: Rule[];
}

​
Group Priority

Groups execute in priority order:
  1. Security Rules (Priority: 0)
  2. Content Policy (Priority: 100)
  3. Community Guidelines (Priority: 200)
  4. Quality Standards (Priority: 300)

​
Advanced Features

​
Rule Variables

Use dynamic values in rules:
VariableDescriptionExample Usage
{{site.id}}Current site IDSite-specific rules
{{user.country}}User’s countryGeo-based filtering
{{time.hour}}Current hour (0-23)Time-based rules
{{media.size}}File size in bytesSize restrictions

​
Rule Functions

Built-in functions for complex logic: 
{
  "conditions": {
    "custom.contains_keywords": {
      "function": "text_contains_any",
      "params": ["spam", "advertisement", "promotion"]
    }
  }
}

​
Rule Templates

Pre-built templates for common scenarios:

COPPA Compliance

Rules for child safety compliance

Brand Safety

Protect brand reputation

Community Standards

Enforce community guidelines

Legal Compliance

Meet regulatory requirements

​
Rule Creation

​
Visual Rule Builder

Create rules without coding:
  1. Select trigger conditions
  2. Add logical operators (AND/OR)
  3. Set confidence thresholds
  4. Choose actions
  5. Test with sample content

​
Code-Based Rules

For complex logic: 
const rule: Rule = {
  name: "Advanced Content Filter",
  evaluate: (content: Content) => {
    const hasViolence = content.ai.violence?.confidence > 0.7;
    const isNewsContext = content.metadata.source === 'news';
    const isVerifiedJournalist = content.user.role === 'journalist';
    
    if (hasViolence && !isNewsContext && !isVerifiedJournalist) {
      return { action: 'reject', reason: 'Violence outside news context' };
    }
    
    return null; // No match
  }
};

​
Testing & Validation

​
Rule Testing

Test rules before deployment: 
// Test rule against sample content
const testResults = await testRule(rule, [
  { url: 'sample1.jpg', expectedAction: 'approve' },
  { url: 'sample2.jpg', expectedAction: 'reject' }
]);

​
Simulation Mode

Run rules without applying actions:
  • See which rules would trigger
  • Review decision explanations
  • Identify conflicts or gaps
  • Fine-tune thresholds

​
Performance Optimization

​
Best Practices

  1. Order by Frequency: Most common rules first
  2. Early Exit: Reject obvious violations quickly
  3. Minimal Conditions: Simpler rules execute faster
  4. Cache Results: Reuse expensive computations

​
Performance Metrics

MetricTargetTypical
Rule Evaluation Time< 10ms2-5ms
Rules per Second> 10,00025,000
Memory Usage< 100MB50MB

​
Monitoring & Analytics

​
Rule Analytics

Track rule effectiveness:
  • Hit Rate: How often each rule triggers
  • Override Rate: Manual overrides of rule decisions
  • Processing Time: Performance of each rule
  • False Positive Rate: Incorrect rejections

​
Dashboard Metrics

interface RuleMetrics {
  ruleId: string;
  triggerCount: number;
  avgProcessingTime: number;
  overrideCount: number;
  lastTriggered: Date;
}

​
Integration Examples

​
E-commerce Platform

{
  "name": "Product Review Filter",
  "conditions": {
    "OR": [
      {"ai.spam.confidence": { ">=": 0.8 }},
      {"text.contains_competitor": true},
      {"user.review_count": { "<": 5 }, "text.contains_url": true}
    ]
  },
  "action": "flag"
}

​
Social Media

{
  "name": "Community Guidelines",
  "conditions": {
    "AND": [
      {"ai.bullying.confidence": { ">=": 0.6 }},
      {"user.previous_violations": { ">": 2 }}
    ]
  },
  "action": "reject",
  "metadata": {
    "notify_user": true,
    "escalate_to_admin": true
  }
}