docker-compose.yml

services:
  # ---------------- KAFKA (KRaft mode — no Zookeeper) ----------------
  # Role: Distributed message broker for real-time chat auditing and logging.
  # apache/kafka:latest is the official ASF image, multi-arch (amd64 + arm64),
  # runs in KRaft mode — no Zookeeper container needed.
  kafka:
    image: apache/kafka:latest
    ports:
      - "9092:9092"
    environment:
      # ── KRaft cluster identity ─────────────────────────────────────
      KAFKA_NODE_ID: 1
      KAFKA_PROCESS_ROLES: "broker,controller"
      KAFKA_CONTROLLER_QUORUM_VOTERS: "1@kafka:9093"
      KAFKA_CONTROLLER_LISTENER_NAMES: "CONTROLLER"
      # ── Listeners ──────────────────────────────────────────────────
      KAFKA_LISTENERS: "PLAINTEXT://:9092,CONTROLLER://:9093"
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: "CONTROLLER:PLAINTEXT,PLAINTEXT:PLAINTEXT"
      KAFKA_ADVERTISED_LISTENERS: "PLAINTEXT://kafka:9092"
      KAFKA_INTER_BROKER_LISTENER_NAME: "PLAINTEXT"
      # ── Reliability ────────────────────────────────────────────────
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 1
      KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR: 1
      KAFKA_TRANSACTION_STATE_LOG_MIN_ISR: 1
      KAFKA_AUTO_CREATE_TOPICS_ENABLE: "true"
    volumes:
      - kafka-data:/var/lib/kafka/data
    healthcheck:
      test: ["CMD", "/opt/kafka/bin/kafka-topics.sh", "--bootstrap-server", "127.0.0.1:9092", "--list"]
      interval: 15s
      timeout: 10s
      retries: 12
    restart: always
    networks:
      - chatbot-net

  # ---------------- KAFKA INIT (Topic Creator) ----------------
  # Role: One-shot container that explicitly creates the chat_logs topic on first boot.
  kafka-init:
    image: apache/kafka:latest
    depends_on:
      kafka:
        condition: service_healthy
    entrypoint: ["/bin/bash", "-c"]
    command:
      - |
        /opt/kafka/bin/kafka-topics.sh --bootstrap-server kafka:9092 --create \
          --if-not-exists --topic chat_logs --partitions 1 --replication-factor 1
        echo "Topic chat_logs is ready."
    networks:
      - chatbot-net

  # ---------------- LUCENE (AI Engine) ----------------
  # Role: Java microservice handling OpenNLP logic and Lucene Fuzzy Search.
  # eclipse-temurin:17-jre-alpine is a multi-arch image (amd64 + arm64).
  lucene:
    build:
      context: "./src/ai-chatbot"
    ports:
      - "4567:4567"
    environment:
      - PORT=4567
      - MODELS_DIR=/app/data/models
      - KB_FILE=/app/data/faq.txt
    volumes:
      - "./src/backend/data:/app/data"
    healthcheck:
      test: ["CMD", "wget", "-qO-", "http://127.0.0.1:4567/health"]
      interval: 15s
      timeout: 5s
      retries: 8
    restart: unless-stopped
    networks:
      - chatbot-net

  # ---------------- BACKEND (Python/FastAPI) ----------------
  # Role: The API gateway and business logic orchestrator.
  # python:3.11-slim is a multi-arch image (amd64 + arm64).
  backend:
    build:
      context: "./src/backend"
    ports:
      - "8000:8000"
    volumes:
      - "chat-logs:/app/logs"
    environment:
      - LUCENE_URL=http://lucene:4567
      - KAFKA_BROKER=kafka:9092
      - REDIS_HOST=redis
      - ENABLE_TRANSLATION=1
      - INTENT_MIN_CONFIDENCE=0.35
      - INTENT_LOW_CONFIDENCE=0.25
      - LUCENE_MIN_SCORE=2.0
    depends_on:
      kafka:
        condition: service_healthy
      lucene:
        condition: service_healthy
      redis:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health', timeout=3)"]
      interval: 15s
      timeout: 5s
      retries: 8
    restart: unless-stopped
    networks:
      - chatbot-net

  # ---------------- REDIS (Session Store) ----------------
  # Role: High-speed in-memory database for managing user sessions and states.
  # redis:alpine is a multi-arch image (amd64 + arm64).
  redis:
    image: redis:alpine
    ports:
      - "6379:6379"
    restart: always
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 15s
      timeout: 5s
      retries: 5
    networks:
      - chatbot-net

  # ---------------- FRONTEND (React) ----------------
  # Role: The User Interface layer (Instagram-style DM UI).
  # node:20.10-slim and nginx:stable-alpine are both multi-arch images.
  frontend:
    build:
      context: "./src/frontend"
    ports:
      - "3000:3000"
    depends_on:
      backend:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "wget", "-qO-", "http://127.0.0.1:3000/"]
      interval: 15s
      timeout: 5s
      retries: 8
    restart: unless-stopped
    networks:
      - chatbot-net

  # ---------------- CONSUMER (Audit Worker) ----------------
  # Role: Background worker that listens to Kafka and persists logs to disk.
  consumer:
    build:
      context: "./src/backend"
    command: ["python", "kafka_consumer.py"]
    environment:
      - KAFKA_BROKER=kafka:9092
    volumes:
      - "chat-logs:/app/logs"
    depends_on:
      kafka:
        condition: service_healthy
    restart: on-failure:5
    networks:
      - chatbot-net

  # ---------------- PROMETHEUS (Metrics Collector) ----------------
  # Role: Scrapes /metrics from the FastAPI backend for time-series storage.
  # prom/prometheus is a multi-arch image (amd64 + arm64).
  prometheus:
    image: prom/prometheus:v2.51.0
    ports:
      - "9090:9090"
    volumes:
      - "./config/monitoring/prometheus.yml:/etc/prometheus/prometheus.yml:ro"
      - prometheus-data:/prometheus
    command:
      - "--config.file=/etc/prometheus/prometheus.yml"
      - "--storage.tsdb.path=/prometheus"
      - "--web.console.libraries=/usr/share/prometheus/console_libraries"
      - "--web.console.templates=/usr/share/prometheus/consoles"
      - "--web.enable-lifecycle"
    depends_on:
      backend:
        condition: service_healthy
    restart: unless-stopped
    networks:
      - chatbot-net

  # ---------------- GRAFANA (Metrics Dashboard) ----------------
  # Role: Visualises Prometheus metrics with pre-loaded dashboards.
  # grafana/grafana is a multi-arch image (amd64 + arm64).
  grafana:
    image: grafana/grafana:10.4.0
    ports:
      - "3001:3000"
    environment:
      - GF_SECURITY_ADMIN_USER=admin
      - GF_SECURITY_ADMIN_PASSWORD=chatbot123
      - GF_USERS_ALLOW_SIGN_UP=false
      - GF_DASHBOARDS_DEFAULT_HOME_DASHBOARD_PATH=/etc/grafana/provisioning/dashboards/grafana_dashboard.json
    volumes:
      - "./config/monitoring/grafana_datasource.yml:/etc/grafana/provisioning/datasources/datasource.yml:ro"
      - "./config/monitoring/grafana_dashboards.yml:/etc/grafana/provisioning/dashboards/dashboards.yml:ro"
      - "./config/monitoring/grafana_dashboard.json:/etc/grafana/provisioning/dashboards/grafana_dashboard.json:ro"
      - grafana-data:/var/lib/grafana
    depends_on:
      - prometheus
    restart: unless-stopped
    networks:
      - chatbot-net

# ---------------- NETWORKS ----------------
networks:
  chatbot-net:
    driver: bridge

# ---------------- VOLUMES ----------------
volumes:
  chat-logs:
  kafka-data:
  prometheus-data:
  grafana-data: