docker-compose.yml

version: "3.9"

services:
  mongodb:
    image: mongo
    container_name: mongodb
    environment:
      - MONGO_INITDB_ROOT_USERNAME=root
      - MONGO_INITDB_ROOT_PASSWORD=pass12345
    ports:
      - 27017:27017
    healthcheck:
      test: echo 'db.stats().ok' | mongosh localhost:27017/test --quiet
      interval: 10s
      timeout: 10s
      retries: 10
    restart: unless-stopped

  loki:
    image: grafana/loki:3.0.0
    command: -config.file=/etc/loki/loki-local.yaml
    user: "0"
    ports:
      - "3101:3100"
    volumes:
      - ./etc/loki-local.yaml:/etc/loki/loki-local.yaml
      - ./data/loki-data:/tmp/loki
    healthcheck:
      test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:3100/ready || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 12

  tempo:
    image: grafana/tempo:2.4.0
    command: ["-config.file=/etc/tempo.yaml"]
    volumes:
      - ./etc/tempo-local.yaml:/etc/tempo.yaml
      - ./data/tempo-data:/tmp/tempo
    restart: unless-stopped
    ports:
      - "3102:3100"   # tempo HTTP API
      - "4317:4317"   # OTLP gRPC
    healthcheck:
      test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:3100/ready || exit 1"]
      interval: 15s
      timeout: 5s
      retries: 12

  # ── LLM inference ────────────────────────────────────────────────────────────
  # Downloads Qwen2.5-1.5B-Instruct Q4_K_M (~1 GB) once into a named volume.
  # Subsequent starts skip the download entirely.
  model-init:
    profiles: ["llm-bundled"]
    image: alpine:3.19
    volumes:
      - llm-models:/models
    entrypoint: ["/bin/sh", "-c"]
    command:
      - |
        apk add --no-cache curl -q
        if [ ! -f /models/qwen2.5-1.5b-instruct-q4_k_m.gguf ]; then
          echo '==> Downloading Qwen2.5-1.5B-Instruct Q4_K_M (~1 GB) — first run only...'
          curl -L --retry 5 --retry-delay 3 --progress-bar \
            -o /models/qwen2.5-1.5b-instruct-q4_k_m.gguf \
            'https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct-GGUF/resolve/main/qwen2.5-1.5b-instruct-q4_k_m.gguf'
          echo '==> Download complete.'
        else
          echo '==> Model already present, skipping download.'
        fi
    restart: "no"

  llama-server:
    profiles: ["llm-bundled"]
    image: ghcr.io/ggerganov/llama.cpp:server
    volumes:
      - llm-models:/models
    command:
      - "-m"
      - "/models/qwen2.5-1.5b-instruct-q4_k_m.gguf"
      - "--host"
      - "0.0.0.0"
      - "--port"
      - "8080"
      - "--ctx-size"
      - "2048"
      - "--alias"
      - "qwen2.5"
      - "-ngl"
      - "0"        # CPU-only; set to 99 if the host has a GPU
    ports:
      - "8082:8080"   # exposed for local debugging; diffy uses internal hostname
    healthcheck:
      test: ["CMD-SHELL", "wget -qO- http://localhost:8080/health 2>/dev/null | grep -q ok"]
      interval: 15s
      timeout: 5s
      retries: 12
      start_period: 30s
    depends_on:
      model-init:
        condition: service_completed_successfully
    restart: unless-stopped
  # ─────────────────────────────────────────────────────────────────────────────

  primary:
    image: diffy/example-service:production

  secondary:
    image: diffy/example-service:production

  candidate:
    image: diffy/example-service:candidate

  diffy:
    image: diffy/diffy:latest
    ports:
      - "8888:8888"
      - "8880:8880"
#    env_file: "./diffy.env"
    environment:
      spring.data.mongodb.authentication-database: admin
      spring.data.mongodb.host: "mongodb"
      spring.data.mongodb.port: 27017
      spring.data.mongodb.username: root
      spring.data.mongodb.password: pass12345
      candidate: "candidate:5000"
      master.primary: "primary:5000"
      master.secondary: "secondary:5000"
      responseMode: "primary"
      service.protocol: "http"
      allowHttpSideEffects: "true"
      serviceName: "Sample Service"
      proxy.port: 8880
      http.port: 8888
      rootUrl: "localhost:8888"
      otel.javaagent.debug: "false"
      otel.metrics.exporter: "none"
      otel.traces.exporter: "otlp"
      otel.exporter.otlp.endpoint: "http://tempo:4317"
      otel.resource.attributes: "service.name=diffy"
      spring.application.name: "diffy"
      logging.level.web: "INFO"
      logging.level.io.opentelemetry: "WARN"
      logging.level.root: "INFO"
      logging.file.name: "/app/logs/diffy.log"
      logging.file.max-size: "10MB"
      logging.file.max-history: 1
      logging.file.clean-history-on-start: "true"
      dockerComposeLocal: "true"
      llm.provider: "${LLM_PROVIDER:-ollama}"
      llm.ollama.url: "${LLM_URL:-http://llama-server:8080}"
      llm.ollama.model: "${LLM_MODEL:-qwen2.5}"
    volumes:
      - ./data/logs:/app/logs
    depends_on:
      - tempo
      - mongodb
      - primary
      - secondary
      - candidate

  promtail:
    image: grafana/promtail:2.9.8
    command: -config.file=/etc/promtail/promtail-local.yaml
    volumes:
      - ./etc/promtail-local.yaml:/etc/promtail/promtail-local.yaml
      - ./data/logs:/app/logs
    depends_on:
      - diffy
      - loki

  prometheus:
    image: prom/prometheus:latest
    volumes:
      - ./etc/prometheus.yaml:/etc/prometheus.yaml
    entrypoint:
      - /bin/prometheus
      - --config.file=/etc/prometheus.yaml
    ports:
      - "9090:9090"
    depends_on:
      - diffy

  grafana:
    image: grafana/grafana:11.2.0
    volumes:
      - ./data/grafana-data/datasources:/etc/grafana/provisioning/datasources
      - ./data/grafana-data/dashboards-provisioning:/etc/grafana/provisioning/dashboards
      - ./data/grafana-data/dashboards:/var/lib/grafana/dashboards
    environment:
      - GF_AUTH_ANONYMOUS_ENABLED=true
      - GF_AUTH_ANONYMOUS_ORG_ROLE=Admin
      - GF_AUTH_DISABLE_LOGIN_FORM=true
    ports:
      - "3000:3000"
    depends_on:
      - prometheus
      - tempo
      - loki

volumes:
  llm-models:   # persists the model across restarts; ~1 GB, downloaded once