KI-Server-Docker: Containerisierung für LLM-Pipelines

version: '3.8'
services:
  # LLM Backend (vLLM)
  llm:
    image: vllm/vllm-openai:latest
    runtime: nvidia
    volumes:
      - model_cache:/root/.cache/huggingface
    ports:
      - "8000:8000"
    command: >
      --model mistralai/Mistral-Large-Instruct-2411
      --max-model-len 32768
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  # Vektordatenbank
  qdrant:
    image: qdrant/qdrant:latest
    ports:
      - "6333:6333"
    volumes:
      - qdrant_data:/qdrant/storage

  # Frontend
  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    ports:
      - "3000:8080"
    environment:
      - OLLAMA_BASE_URL=http://llm:8000

volumes:
  model_cache:
  qdrant_data:

# 1. Server vorbereiten
nvidia-smi  # GPU erkannt?
docker --version  # Docker installiert?

# 2. Pipeline starten
docker compose up -d

# 3. Status prüfen
docker compose ps

# 4. Logs ansehen
docker compose logs -f llm

# Vektordatenbank sichern
docker exec qdrant tar czf /backup/qdrant-$(date +%Y%m%d).tar.gz /qdrant/storage

# Docker Compose Config sichern
cp docker-compose.yml /backup/docker-compose-$(date +%Y%m%d).yml