services:
  llama.cpp:
    image: ghcr.io/ggerganov/llama.cpp:server
    command: -m /models/tinyllama-1.1b-chat-v1.0.Q2_K.gguf -c 512 --host 0.0.0.0 --port 8080
    ports:
    - "8080:8080"
    volumes:
    - ./data:/models
    restart: unless-stopped