n8n Docker Compose Stack

A complete Docker Compose setup for running n8n automation workflows with integrated AI services and document processing capabilities.

🚀 Quick Start

Prerequisites

• Docker and Docker Compose installed
• At least 4GB RAM available (for AI models)
• 10GB free disk space

1. Clone and Start

git clone <repository-url>
cd n8n-docker-compose

Start the default (basic) set (ytdlp, whisper, n8n):

docker-compose up -d --build

Start the full set (all services):

docker-compose --profile full up -d --build

2. Access n8n

• Open http://localhost:5678
• Login with:
  • Username: admin
  • Password: password

3. Import Sample Workflow

1. In n8n, go to Workflows → Import from File
2. Upload workflows/Text File to Elasticsearch Indexer - Working Version.json
3. Activate the workflow

🎯 What is this repository?

This repository provides a ready-to-use Docker Compose stack that includes:

• n8n - No-code automation platform for building workflows
• Elasticsearch - Search and analytics engine for document indexing
• Ollama - Local LLM inference server for AI capabilities
• LaBSE - Multilingual text embeddings service (109 languages supported)
• YT-DLP - YouTube and video platform downloader with REST API
• Whisper - Audio transcription service using OpenAI Whisper (faster-whisper)

🎯 Why use this stack?

This setup is ideal for:

• Document Processing & Search: Upload text files, automatically split them into paragraphs, and index them in Elasticsearch for semantic search
• AI-Powered Workflows: Leverage local LLM models through Ollama for text processing, analysis, and generation
• Multilingual Text Analysis: Use LaBSE embeddings for cross-language semantic similarity and search
• Audio Transcription: Convert audio files to text using state-of-the-art Whisper models
• No-Code Automation: Build complex workflows without programming using n8n's visual interface
• Privacy-First AI: All AI processing happens locally — no data sent to external APIs

📋 Services Overview

Service	Port	Purpose	Access
n8n	5678	Workflow automation platform	http://localhost:5678
Elasticsearch	-	Document search and analytics	Internal (via Docker network)
Ollama	-	Local LLM inference server	Internal (via Docker network)
LaBSE	-	Multilingual embeddings API	Internal (via Docker network)
YT-DLP	-	YouTube video/audio downloader	Internal (via Docker network)
Whisper	-	Audio transcription service	Internal (via Docker network)

Note: Ports for services other than n8n are not forwarded to localhost to avoid conflicts. They are accessible from n8n or other containers via the Docker network.

🎬 Ready-to-Use Workflows

YouTube Video Transcription (Manual)

Automatically transcribe YouTube videos to text:

1. Open n8n: http://localhost:5678 (login: admin / password)
2. Find workflow: "YouTube Video Transcription (Manual)"
3. Edit URL: Open the "Enter YouTube URL" node and paste your YouTube link
4. Run: Click "Execute workflow"
5. Get result: View the transcription in the "Format Response" node

📖 Full instructions: See workflows/QUICK_START.md

Features:

• ✅ Simple form-based input
• ✅ Automatic audio extraction
• ✅ Multi-language support (auto-detect or specify)
• ✅ Timestamped segments
• ✅ Multiple Whisper models (tiny to large)

📊 Sample Workflow: Text File Indexer

The included workflow demonstrates the stack's capabilities:

1. File Upload: Web form for uploading text files
2. Text Processing: Automatically splits content into paragraphs
3. Elasticsearch Indexing: Stores processed documents with metadata
4. Batch Processing: Handles large files efficiently

How to use:

1. Activate the "Text File to Elasticsearch Indexer" workflow
2. Open the webhook URL provided by the form trigger
3. Upload a .txt file
4. Documents will be automatically indexed in Elasticsearch

🔧 Service Details

n8n Configuration

• Basic authentication enabled
• Persistent data storage
• Connected to other services via Docker network
• Production-oriented settings for local development

Elasticsearch

• Single-node setup
• No authentication (for local development)
• Persistent data volume
• Optimized for document storage

Ollama

• Ready for LLM model installation
• Persistent model storage
• GPU acceleration supported if available

LaBSE Embeddings

• FastAPI-based service
• Supports 109 languages
• 768-dimensional embeddings
• Batch processing capable

YT-DLP Service

• Download videos from YouTube and 1000+ sites
• Extract audio in MP3 format
• Return video metadata and transcripts
• REST API for easy integration

Whisper Transcription

• Audio-to-text transcription using faster-whisper
• Supports multiple audio formats (MP3, WAV, M4A, OGG, FLAC, WebM)
• Automatic language detection
• Multiple model sizes (tiny, base, small, medium, large)
• Timestamp generation for each segment
• Optional translation to English

📚 Usage Examples

Generate Embeddings (requires --profile full)

curl -X POST http://localhost:8080/embeddings \
  -H "Content-Type: application/json" \
  -d '{"texts": ["Hello world", "Привет мир", "Hola mundo"]}'

Search Elasticsearch (requires --profile full)

curl -X GET "http://localhost:9200/documents/_search?q=your_search_term"

Install Ollama Models (requires --profile full)

docker exec -it ollama ollama pull llama2

Download YouTube Video (available in the basic set)

curl -X POST http://localhost:8081/download \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.youtube.com/watch?v=VIDEO_ID", "format": "video", "quality": "best"}'

Download YouTube Audio (MP3) (available in the basic set)

curl -X POST http://localhost:8081/download \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.youtube.com/watch?v=VIDEO_ID", "format": "audio"}'

Transcribe Audio File (available in the basic set)

# Basic transcription
curl -X POST http://localhost:8082/transcribe \
  -F "file=@audio.mp3"

# With specific model and language
curl -X POST http://localhost:8082/transcribe \
  -F "file=@audio.mp3" \
  -F "model=small" \
  -F "language=ru"

# Translate to English
curl -X POST http://localhost:8082/transcribe \
  -F "file=@audio.mp3" \
  -F "translate=true"

🔍 API Documentation

• LaBSE API: http://localhost:8080/docs (requires --profile full)
• YT-DLP API: See docker/ytdlp/README.md for full API documentation
• Whisper API: See docker/whisper/README.md for full API documentation
• Elasticsearch: http://localhost:9200 (requires --profile full)
• n8n Webhooks: Available through workflow triggers

💾 Data Persistence

All data is persisted in Docker volumes:

• n8n_data - n8n workflows and settings
• esdata - Elasticsearch indices
• ollama_data - Downloaded LLM models
• labse_cache - Cached embedding models
• ytdlp_downloads - Downloaded videos and audio files
• ytdlp_cache - yt-dlp cache
• whisper_cache - Whisper models cache

🚀 Advanced Usage

Custom Workflows

Create workflows that combine:

• File processing with LaBSE embeddings
• AI text generation with Ollama
• Document search with Elasticsearch
• Web scraping and data transformation
• Download YouTube videos with YT-DLP and transcribe with Whisper
• Audio transcription and text analysis with Ollama

Scaling

• Add more Elasticsearch nodes for clustering
• Use external GPU for Ollama acceleration
• Deploy behind reverse proxy for production

🛠 Troubleshooting

Common Issues

Services not starting?

• Check Docker logs: docker-compose logs [service-name]
• Ensure ports are not in use
• Verify sufficient disk space and RAM

n8n workflows failing?

• Check service connectivity between containers
• Verify Elasticsearch is ready before running workflows
• Check webhook URLs and credentials

LaBSE model loading slowly?

• First request downloads ~1.8GB model
• Subsequent requests are fast
• Model is cached in volume

📝 Default Credentials

• n8n: admin / password
• Elasticsearch: No authentication
• Other services: No authentication (local development)

🤝 Contributing

Feel free to:

• Add new workflows to the workflows/ directory
• Improve service configurations
• Add new AI services or tools
• Submit issues and pull requests

📄 License

This project is provided as-is for educational and development purposes.