backstory/docker-compose.yml

services:
  backstory:
    build:
      context: .
      dockerfile: Dockerfile
      target: backstory
    container_name: backstory
    image: backstory
    restart: "no"
    env_file:
      - .env
    environment:
      - PRODUCTION=0
      - MODEL_NAME=${MODEL_NAME:-qwen2.5:3b}
    devices:
      - /dev/dri:/dev/dri
    depends_on:
      - ollama
    networks:
      - internal
    ports:
      - 8912:8911 # Flask React server
      - 3000:3000 # REACT expo while developing frontend
    volumes:
      - ./cache:/root/.cache                # Persist all models and GPU kernel cache
      - ./sessions:/opt/backstory/sessions:rw    # Persist sessions
      - ./chromadb:/opt/backstory/chromadb:rw    # Persist ChromaDB
      - ./docs:/opt/backstory/docs:ro            # Live mount of RAG content
      - ./src:/opt/backstory/src:rw              # Live mount server src 
      - ./frontend:/opt/backstory/frontend:rw    # Live mount frontend src 
    cap_add: # used for running ze-monitor within container
      - CAP_DAC_READ_SEARCH # Bypass all filesystem read access checks
      - CAP_PERFMON         # Access to perf_events (vs. overloaded CAP_SYS_ADMIN)
      - CAP_SYS_PTRACE      # PTRACE_MODE_READ_REALCREDS ptrace access mode check

  backstory-prod:
    build:
      context: .
      dockerfile: Dockerfile
      target: backstory
    image: backstory
    container_name: backstory-prod
    restart: "always"
    env_file:
      - .env
    environment:
      - PRODUCTION=1
      - MODEL_NAME=${MODEL_NAME:-qwen2.5:7b}
    devices:
      - /dev/dri:/dev/dri
    depends_on:
      - ollama
    networks:
      - internal
    ports:
      - 8911:8911 # Flask React server
    volumes:
      - ./cache:/root/.cache                # Persist all models and GPU kernel cache
      - ./chromadb-prod:/opt/backstory/chromadb:rw    # Persist ChromaDB
      - ./sessions:/opt/backstory/sessions:rw    # Persist sessions
      - ./docs:/opt/backstory/docs:ro            # Live mount of RAG content
      - ./frontend:/opt/backstory/frontend:rw    # Live mount frontend src 
    cap_add: # used for running ze-monitor within container
      - CAP_DAC_READ_SEARCH # Bypass all filesystem read access checks
      - CAP_PERFMON         # Access to perf_events (vs. overloaded CAP_SYS_ADMIN)
      - CAP_SYS_PTRACE      # PTRACE_MODE_READ_REALCREDS ptrace access mode check

  ollama:
    build:
      context: .
      dockerfile: Dockerfile
      target: ollama
    image: ollama
    container_name: ollama
    restart: "always"
    env_file:
      - .env
    environment:
      - OLLAMA_HOST=0.0.0.0
      - ONEAPI_DEVICE_SELECTOR=level_zero:0
    devices:
      - /dev/dri:/dev/dri
    ports:
      - 11434:11434 # ollama serve port
    networks:
      - internal
    volumes:
      - ./cache:/root/.cache   # Cache hub models and neo_compiler_cache
      - ./ollama:/root/.ollama # Cache the ollama models
    cap_add: # used for running ze-monitor within container
      - CAP_DAC_READ_SEARCH # Bypass all filesystem read access checks
      - CAP_PERFMON         # Access to perf_events (vs. overloaded CAP_SYS_ADMIN)
      - CAP_SYS_PTRACE      # PTRACE_MODE_READ_REALCREDS ptrace access mode check

  jupyter:
    build:
      context: .
      dockerfile: Dockerfile
      target: jupyter
    image: jupyter
    container_name: jupyter
    restart: "always"
    env_file:
      - .env
    devices:
      - /dev/dri:/dev/dri
    depends_on:
      - ollama
      - miniircd
    ports:
    - 8888:8888 # Jupyter Notebook
    - 60673:60673 # Gradio
    networks:
      - internal
    volumes:
      - ./jupyter:/opt/jupyter:rw
      - ./cache:/root/.cache
    deploy:
      resources:
        limits:
          memory: "0"  # No memory limit (Docker treats 0 as unlimited)
        reservations:
          memory: "0"  # No reserved memory (optional)
    ulimits:
      memlock: -1  # Prevents memory from being locked
    #oom_kill_disable: true  # Prevents OOM killer from killing the container
    cap_add: # used for running ze-monitor within container
      - CAP_DAC_READ_SEARCH # Bypass all filesystem read access checks
      - CAP_PERFMON         # Access to perf_events (vs. overloaded CAP_SYS_ADMIN)
      - CAP_SYS_PTRACE      # PTRACE_MODE_READ_REALCREDS ptrace access mode check

  miniircd:
    build:
      context: .
      dockerfile: Dockerfile
      target: miniircd
    image: miniircd
    container_name: miniircd
    restart: "no"
    env_file:
      - .env
    devices:
      - /dev/dri:/dev/dri
    ports:
    - 6667:6667 # IRC
    networks:
      - internal
    volumes:
      - ./cache:/root/.cache

networks:
  internal:
    driver: bridge