PeerTube Plugin: German AI Moderation

AI-powered comment moderation plugin for PeerTube using two-stage German toxicity detection with ml6team DistilBERT and deepset BERT models.

Features

• Dual-Model System:
  • Primary: ml6team/distilbert-base-german-cased-toxic-comments - Fast general toxicity detection (67M, optimized for comments)
  • Secondary: deepset/bert-base-german-cased-hatespeech-GermEval18Coarse - Specialized hate speech detection
• German Language Detection: Automatically detects comment language using langdetect
• Two-Level Filtering:
  • General toxicity threshold (configurable, default: 0.7)
  • Hate speech threshold (configurable, default: 0.5)
• Non-German Comments: Configurable action for comments in other languages (moderation queue or block)
• Fail-Open Strategy: Comments are allowed if AI service is unavailable (prevents site breakage)
• Moderation Interface: Admin panel for reviewing and approving blocked comments

Architecture

The plugin consists of two components:

1. PeerTube Plugin (peertube-plugin-aimod/) - PeerTube plugin that intercepts comments and sends them to AI service
2. AI Moderation Service (ai-moderation/) - FastAPI service that runs ML models for toxicity detection

Requirements

• PeerTube >= 7.0.0
• Docker and Docker Compose
• AI moderation service running in Docker network
• ~2GB disk space for ML models (cached in /srv/ai-model-cache)

Installation

1. Install PeerTube Plugin

Install via PeerTube admin panel:

• Go to Administration → Plugins/Themes
• Click Install a plugin/the theme
• Enter package name: peertube-plugin-german-ai-mod
• Or upload ZIP file from releases

Or install via npm:

npm install peertube-plugin-german-ai-mod

2. Setup AI Moderation Service

Add the AI moderation service to your docker-compose.yml:

services:
  ai-moderator:
    build: ./ai-moderation
    container_name: ai-moderator
    restart: unless-stopped
    networks: [web]
    environment:
      - HF_HOME=/models
      - TZ=Europe/Berlin
    volumes:
      - /srv/ai-model-cache:/models
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
      interval: 30s
      timeout: 5s
      start_period: 40s
      retries: 3

Or use the provided docker-compose-ai-mod.yml file and merge it with your existing docker-compose configuration.

Important: Make sure the AI service is in the same Docker network as PeerTube (usually web network).

3. Start Services

docker-compose up -d ai-moderator

The first startup will download ML models (~2GB), which may take several minutes.

Configuration

After installation, configure the plugin in PeerTube admin panel:

• AI Endpoint URL: URL of the AI moderation service (default: http://ai-moderator:8000/analyze)
  • Must be accessible from PeerTube container
  • Use Docker service name for internal network communication
• Toxizitäts-Schwelle (0–1): Threshold for general toxicity detection (default: 0.7)
• Hate Speech Schwelle (0–1): Threshold for hate speech detection (default: 0.5)
• Aktion für nicht-deutsche Kommentare:
  • Zur Moderation zurückhalten - Send to moderation queue (default)
  • Blockieren - Block immediately

Recommended Thresholds

• Moderate (default): Toxicity 0.7, Hate Speech 0.5 - Blocks only explicit insults
• Strict: Toxicity 0.5, Hate Speech 0.4 - Blocks more aggressive comments
• Soft: Toxicity 0.8, Hate Speech 0.6 - Blocks only very explicit insults

Moderation Interface

Access the moderation interface at:

/admin/plugins/german-ai-mod/moderation

Features:

• View all blocked comments
• Filter by reason (toxic, non-German), status (pending, approved), or video
• Approve comments (unblock and publish)
• Delete comments permanently
• View toxicity scores and language detection results

How It Works

1. When a comment is created, the plugin sends it to the AI service
2. AI service detects the language using langdetect
3. For non-German comments: sends to moderation or blocks (based on settings)
4. For German comments:
   • Primary model checks general toxicity
   • Secondary model checks for hate speech
   • Comment is blocked if score >= toxicity_threshold OR hate_score >= hate_threshold
5. Toxic/hate speech comments are automatically blocked and saved to database
6. Moderators can review and approve comments via admin interface

API Endpoints

The plugin provides REST API endpoints for moderation:

• GET /plugins/german-ai-mod/router/blocked-comments - List blocked comments
• POST /plugins/german-ai-mod/router/blocked-comments/:id/approve - Approve a comment
• DELETE /plugins/german-ai-mod/router/blocked-comments/:id - Delete a comment
• GET /plugins/german-ai-mod/router/my-videos - Get user's videos for filtering

AI Service API

The AI service provides a POST endpoint /analyze that accepts:

{
  "text": "comment text here"
}

And returns:

{
  "toxic": false,
  "score": 0.15,
  "language": "de",
  "is_german": true,
  "requires_moderation": false,
  "hate_score": 0.05
}

Where:

• toxic: boolean - whether comment should be blocked (based on thresholds)
• score: float (0-1) - general toxicity score from primary model
• hate_score: float (0-1) - hate speech score from secondary model
• language: string - detected language code
• is_german: boolean - whether comment is in German
• requires_moderation: boolean - whether comment requires manual review

Hooks Used

• filter:api.video-thread.create.accept.result - Main comments
• filter:api.video-comment-reply.create.accept.result - Replies
• filter:activity-pub.remote-video-comment.create.accept.result - Federated comments

Models Used

• Primary: ml6team/distilbert-base-german-cased-toxic-comments

  • Lightweight (67M parameters)
  • Trained on 5 German toxicity datasets
  • Optimized for comment toxicity detection

• Secondary: deepset/bert-base-german-cased-hatespeech-GermEval18Coarse

  • Specialized for hate speech detection
  • Trained on GermEval18 dataset
  • Coarse-grained classification

Performance

• First request: ~5-10 seconds (model loading)
• Subsequent requests: ~200-500ms per comment
• Models are cached in /srv/ai-model-cache volume
• Health check endpoint: /health

Troubleshooting

AI service not responding

1. Check if container is running: docker ps | grep ai-moderator
2. Check logs: docker logs ai-moderator
3. Verify network connectivity: docker exec peertube curl http://ai-moderator:8000/health
4. Check endpoint URL in plugin settings

Models not loading

1. Check disk space: df -h /srv/ai-model-cache
2. Check logs for download errors: docker logs ai-moderator
3. Verify HuggingFace access (models are public, no token needed)

Comments not being blocked

1. Check plugin logs in PeerTube: docker logs peertube | grep "AI Mod"
2. Verify thresholds in plugin settings
3. Test AI service directly: curl -X POST http://ai-moderator:8000/analyze -H "Content-Type: application/json" -d '{"text":"test comment"}'

Development

Building AI Service

cd ai-moderation
docker build -t ai-moderator .

Testing AI Service Locally

cd ai-moderation
pip install -r requirements.txt
uvicorn main:app --host 0.0.0.0 --port 8000

Plugin Development

cd peertube-plugin-aimod
npm install

License

AGPL-3.0

Author

yarkolife

Repository

https://github.com/yarkolife/peertube-plugin-german-ai-mod