# Docker Compose for ollama-d development, integration testing, and the
# Open WebUI chatbot dashboard.
#
# Services:
#   ollama   — Ollama inference server (GPU pass-through optional).
#   webui    — Open WebUI: chatbot frontend + realtime system/model dashboard.
#   test     — D SDK build + integration test runner (one-shot).
#
# Usage:
#   # Start Ollama + Open WebUI only (chatbot at http://localhost:3000):
#   docker compose up ollama webui
#
#   # Run full integration tests (builds SDK, runs samples against Ollama):
#   docker compose up --build
#
#   # Tear down and remove volumes:
#   docker compose down -v

services:

  # ── Ollama inference server ────────────────────────────────────────────────
  ollama:
    image: ollama/ollama:latest
    ports:
      - "11434:11434"
    volumes:
      - ollama_data:/root/.ollama
    # Uncomment for NVIDIA GPU pass-through:
    # deploy:
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: all
    #           capabilities: [gpu]
    healthcheck:
      test: ["CMD", "curl", "-sf", "http://localhost:11434/api/version"]
      interval: 10s
      timeout: 5s
      retries: 10
      start_period: 15s
    restart: unless-stopped

  # ── Open WebUI — chatbot + realtime dashboard ──────────────────────────────
  # Provides:
  #   • Full chat interface with model selection and history
  #   • Admin dashboard: realtime CPU/memory usage, GPU stats, model status
  #   • Model management: pull/delete models from the UI
  #   • OpenAI-compatible API passthrough
  #
  # Access at:  http://localhost:3000
  # Admin panel: http://localhost:3000/admin/settings
  webui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: open-webui
    ports:
      - "3000:8080"
    volumes:
      - open_webui_data:/app/backend/data
    environment:
      OLLAMA_BASE_URL: http://ollama:11434
      # Disable login prompt for local development:
      WEBUI_AUTH: "False"
    depends_on:
      ollama:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "curl", "-sf", "http://localhost:8080/health"]
      interval: 15s
      timeout: 5s
      retries: 5
      start_period: 30s
    restart: unless-stopped

  # ── D SDK build + integration test runner ─────────────────────────────────
  test:
    build:
      context: .
      target: integration
    depends_on:
      ollama:
        condition: service_healthy
    environment:
      OLLAMA_HOST: http://ollama:11434
    entrypoint: ["sh", "-c"]
    command:
      - |
        apt-get install -y --no-install-recommends curl > /dev/null 2>&1 || true
        until curl -sf $$OLLAMA_HOST/api/version; do
          echo "Waiting for Ollama..."; sleep 3
        done
        echo "Pulling model..."
        curl -s -X POST $$OLLAMA_HOST/api/pull \
          -d '{"name":"llama3.1:8b","stream":false}' > /dev/null
        echo "Running simple sample (includes streaming + search)..."
        dub run -b release :simple
        echo "Running coder sample (streaming)..."
        dub run -b release :coder -- \
          --prompt "Write a D function to compute fibonacci numbers" \
          --model llama3.1:8b \
          --host $$OLLAMA_HOST \
          --verbose

volumes:
  ollama_data:
  open_webui_data: