Self-Hosted AI Stack: Ollama, n8n, Open WebUI & Co.

┌─────────────────────────────────────┐
│   Web-UI (Open WebUI, Chatbot)      │  Port 8080
├─────────────────────────────────────┤
│   LLM Inference (Ollama, vLLM)      │  Port 11434
├─────────────────────────────────────┤
│   Workflows (n8n, Activepieces)     │  Port 5678
├─────────────────────────────────────┤
│   Knowledge (open-notebook, RAG)    │  Port 8000
├─────────────────────────────────────┤
│   Vector DB (ChromaDB, Qdrant)      │  Port 6333
├─────────────────────────────────────┤
│   Monitoring (Grafana, Prometheus)  │  Port 3000
├─────────────────────────────────────┤
│   Database (PostgreSQL, SurrealDB)  │  Port 5432
├─────────────────────────────────────┤
│   Hardware (GPU, CPU, RAM)          │  Server
└─────────────────────────────────────┘

brew install ollama
ollama run llama2  # Startet LLM Download + Chat

curl https://ollama.ai/install.sh | sh
sudo systemctl start ollama

ollama run neural-chat  # Schnelleres Modell (7B)

# Lokale Chat-API
curl http://localhost:11434/api/chat -d '{
  "model": "neural-chat",
  "messages": [{"role": "user", "content": "Hallo"}]
}'

# In n8n, Open WebUI, anderen Apps direkt verwenden

pip install vllm

vllm serve meta-llama/Llama-2-7b-hf \
  --max-model-len 4096 \
  --port 8000 \
  --tensor-parallel-size 2  # Auf 2 GPUs verteilt

version: '3'
services:
  open-webui:
    image: ghcr.io/open-webui/open-webui:latest
    ports:
      - "8080:8080"
    volumes:
      - open-webui-data:/app/backend/data
    environment:
      OLLAMA_API_BASE_URL: http://ollama:11434
    depends_on:
      - ollama

  ollama:
    image: ollama/ollama:latest
    ports:
      - "11434:11434"
    volumes:
      - ollama-data:/root/.ollama
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

volumes:
  open-webui-data:
  ollama-data:

docker-compose up -d
# Offen in Browser: http://localhost:8080

version: '3'
services:
  n8n:
    image: n8nio/n8n:latest
    ports:
      - "5678:5678"
    environment:
      N8N_HOST: localhost
      N8N_PORT: 5678
      N8N_PROTOCOL: http
      N8N_SECURE_COOKIE: "false"
      DB_TYPE: postgres
      DB_POSTGRESDB_HOST: postgres
      DB_POSTGRESDB_USER: n8n
      DB_POSTGRESDB_PASSWORD: n8n_password
      DB_POSTGRESDB_DATABASE: n8n
    volumes:
      - n8n-data:/home/node/.n8n
    depends_on:
      - postgres

  postgres:
    image: postgres:13
    environment:
      POSTGRES_USER: n8n
      POSTGRES_PASSWORD: n8n_password
      POSTGRES_DB: n8n
    volumes:
      - postgres-data:/var/lib/postgresql/data

volumes:
  n8n-data:
  postgres-data:

docker-compose up -d
# http://localhost:5678

n8n Node: HTTP Request
  URL: http://ollama:11434/api/chat
  Method: POST
  Body: {
    "model": "neural-chat",
    "messages": [{"role": "user", "content": "{{ $json.input }}"}]
  }

docker run -d \
  -p 80:80 \
  -e FRONTEND_URL=http://localhost \
  -e BACKEND_URL=http://localhost/api \
  -v activepieces-data:/root/.activepieces \
  activepieces/activepieces:latest

service:
  open-notebook:
    image: your-registry/open-notebook:latest
    ports:
      - "8000:8000"
    environment:
      OLLAMA_API_BASE: http://ollama:11434
      SURREALDB_URL: surreal://surrealdb:8000
    depends_on:
      - ollama
      - surrealdb

import chromadb

client = chromadb.Client()
collection = client.create_collection("documents")

# Add documents
collection.add(
    documents=["Kundenvertrag von 2024", "Budget Excel"],
    metadatas=[{"source": "legal"}, {"source": "finance"}],
    ids=["doc1", "doc2"]
)

# Search
results = collection.query(
    query_texts=["Zahlungsbedingungen"],
    n_results=2
)

docker run -p 8000:8000 chromadb/chroma:latest

qdrant:
  image: qdrant/qdrant:latest
  ports:
    - "6333:6333"
  volumes:
    - qdrant-data:/qdrant/storage

docker run -d \
  -e POSTGRES_PASSWORD=password \
  -p 5432:5432 \
  postgres:15

docker run -d \
  -p 8000:8000 \
  surrealdb/surrealdb:latest start

docker run -d -p 9090:9090 prom/prometheus
docker run -d -p 3000:3000 grafana/grafana
docker run -d -p 3001:3001 louislam/uptime-kuma

Internet ← Cloudflare Tunnel ← Reverse Proxy (Nginx)
                                      ↓
         ┌────────────────┬───────────┼───────────┐
         ↓                ↓           ↓           ↓
     n8n:5678    Open WebUI:8080  Ollama:11434  Monitoring:3000
         ↓                ↓           ↓           ↓
         └────────────────┴─────────┬─────────────┘
                                   ↓
                        PostgreSQL 5432
                        SurrealDB 8000
                        ChromaDB 8100

#!/bin/bash
# backup.sh

BACKUP_DIR="/backups/$(date +%Y-%m-%d)"
mkdir -p $BACKUP_DIR

# n8n Workflows (JSON)
docker exec n8n npm --prefix . run build 2>&1 | grep -i workflow > $BACKUP_DIR/n8n-workflows.json

# PostgreSQL
docker exec postgres pg_dump -U n8n n8n > $BACKUP_DIR/n8n-db.sql

# SurrealDB
docker exec surrealdb surreal export 'surreal://localhost:8000' > $BACKUP_DIR/surreal.sql

# Open WebUI Chats
docker cp open-webui:/app/backend/data/webui.db $BACKUP_DIR/

# Compress
tar -czf $BACKUP_DIR.tar.gz $BACKUP_DIR/
aws s3 cp $BACKUP_DIR.tar.gz s3://your-bucket/backups/

0 2 * * * /root/backup.sh >> /root/backup.log 2>&1

version: '3.8'

services:
  ollama:
    image: ollama/ollama:latest
    ports:
      - "11434:11434"
    volumes:
      - ollama-data:/root/.ollama
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
      interval: 30s
      timeout: 10s
      retries: 3

  open-webui:
    image: ghcr.io/open-webui/open-webui:latest
    ports:
      - "8080:8080"
    volumes:
      - open-webui-data:/app/backend/data
    environment:
      OLLAMA_API_BASE_URL: http://ollama:11434
    depends_on:
      ollama:
        condition: service_healthy

  n8n:
    image: n8nio/n8n:latest
    ports:
      - "5678:5678"
    volumes:
      - n8n-data:/home/node/.n8n
    environment:
      N8N_HOST: localhost
      N8N_PORT: 5678
      DB_TYPE: postgres
      DB_POSTGRESDB_HOST: postgres
      DB_POSTGRESDB_USER: n8n
      DB_POSTGRESDB_PASSWORD: ${DB_PASSWORD}
      DB_POSTGRESDB_DATABASE: n8n
    depends_on:
      - postgres

  postgres:
    image: postgres:15
    volumes:
      - postgres-data:/var/lib/postgresql/data
    environment:
      POSTGRES_USER: n8n
      POSTGRES_PASSWORD: ${DB_PASSWORD}
      POSTGRES_DB: n8n
    ports:
      - "5432:5432"

  prometheus:
    image: prom/prometheus:latest
    ports:
      - "9090:9090"
    volumes:
      - ./prometheus.yml:/etc/prometheus/prometheus.yml
      - prometheus-data:/prometheus

  grafana:
    image: grafana/grafana:latest
    ports:
      - "3000:3000"
    volumes:
      - grafana-data:/var/lib/grafana
    environment:
      GF_SECURITY_ADMIN_PASSWORD: ${GRAFANA_PASSWORD}

  uptime-kuma:
    image: louislam/uptime-kuma:latest
    ports:
      - "3001:3001"
    volumes:
      - uptime-kuma-data:/app/data

volumes:
  ollama-data:
  open-webui-data:
  n8n-data:
  postgres-data:
  prometheus-data:
  grafana-data:
  uptime-kuma-data:

export DB_PASSWORD=secure_password_here
export GRAFANA_PASSWORD=another_secure_password
docker-compose up -d