Self-Hosted AI Starter Kit — Architecture & Components
Included Services
| Service | Purpose | Port |
|---|---|---|
| n8n | Low-code workflow automation with 400+ integrations and AI components | 5678 |
| Ollama | Run LLMs locally (Llama 3, Mistral, CodeLlama, etc.) | 11434 |
| Qdrant | High-performance vector database for embeddings and RAG | 6333 |
| PostgreSQL | Persistent data storage for n8n workflows and credentials | 5432 |
What You Can Build
- RAG Chatbots — Upload documents, embed with Ollama, store in Qdrant, query via n8n chat interface
- AI Workflow Automation — Trigger AI tasks from emails, webhooks, schedules, or 400+ app integrations
- Document Processing — Extract, summarize, and classify documents using local LLMs
- Code Generation — Use CodeLlama for automated code review and generation workflows
- Data Analysis — Connect databases to LLMs for natural language data queries
GPU Support
| Platform | Command | Notes |
|---|---|---|
| CPU only | docker compose --profile cpu up |
Slower but works everywhere |
| NVIDIA GPU | docker compose --profile gpu-nvidia up |
Requires NVIDIA Docker runtime |
| AMD GPU | docker compose --profile gpu-amd up |
Linux only |
| Apple Silicon | docker compose up |
CPU mode, or use local Ollama |
Included Workflow Templates
The starter kit comes with pre-built n8n workflow templates:
- AI chatbot with document RAG
- Email classification and auto-response
- Webhook-triggered AI processing
Configuration
Key .env settings:
# n8n settings
N8N_ENCRYPTION_KEY=your-encryption-key
N8N_USER_MANAGEMENT_JWT_SECRET=your-jwt-secret
# Ollama model (pulled on first start)
OLLAMA_MODEL=llama3
# PostgreSQL
POSTGRES_PASSWORD=your-passwordFAQ
Q: What is the Self-Hosted AI Starter Kit? A: A Docker Compose template by n8n that sets up a complete local AI environment with n8n (workflow automation), Ollama (local LLMs), Qdrant (vector DB), and PostgreSQL in one command.
Q: Is it free? A: Yes, fully open-source under Apache 2.0. All included services are free and self-hosted.
Q: What hardware do I need? A: Minimum 8GB RAM for CPU mode. For GPU acceleration, an NVIDIA or AMD GPU with Docker support. Apple Silicon Macs work with CPU mode.