services:
  redis:
    image: redis:7.2-alpine
    ports:
      - "6379:6379"
    volumes:
      - redis_data:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 5

  postgres_db:
    image: postgres:15-alpine
    ports:
      - "5433:5432" # Host port 5433 to avoid conflict if local Postgres runs on 5432
    environment:
      POSTGRES_USER: ${POSTGRES_USER:-user}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-password}
      POSTGRES_DB: ${POSTGRES_DB:-knowledge_assistant_db}
    volumes:
      - postgres_data:/var/lib/postgresql/data
      - ./init-db.sh:/docker-entrypoint-initdb.d/init-db.sh # Optional: script to init DB on first run
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-user} -d ${POSTGRES_DB:-knowledge_assistant_db}"]
      interval: 10s
      timeout: 5s
      retries: 5

  vector_db: # ChromaDB as an example
    image: chromadb/chroma:0.5.0 # Use a specific version
    ports:
      - "8001:8000" # Host port 8001 to avoid conflict
    volumes:
      - chroma_data:/chroma/chroma # Persist Chroma data
    environment:
      - IS_PERSISTENT=TRUE
      - CHROMA_SERVER_HOST=0.0.0.0 # Listen on all interfaces inside container
      - CHROMA_SERVER_HTTP_PORT=8000
      - ANONYMIZED_TELEMETRY=FALSE
    # command: uvicorn chromadb.app:app --workers 1 --host 0.0.0.0 --port 8000 --log-config chromadb/log_config.yml
    # Healthcheck for Chroma might be tricky, often check HTTP endpoint
    # healthcheck:
    #   test: ["CMD", "curl", "-f", "http://localhost:8000/api/v1/heartbeat"]
    #   interval: 30s
    #   timeout: 10s
    #   retries: 3

  api:
    build:
      context: .
      dockerfile: Dockerfile.api
    ports:
      - "${API_PORT:-8080}:${API_PORT:-8080}"
    depends_on:
      redis:
        condition: service_healthy
      postgres_db:
        condition: service_healthy
      vector_db: # Worker needs vector_db, API might not directly but good to have it up
        condition: service_started # ChromaDB takes a bit to start, using service_started
    environment:
      - LOG_LEVEL=${LOG_LEVEL:-INFO}
      - CELERY_BROKER_URL=${CELERY_BROKER_URL:-redis://redis:6379/0}
      - DATABASE_URL=${DATABASE_URL:-postgresql://${POSTGRES_USER:-user}:${POSTGRES_PASSWORD:-password}@postgres_db:5432/${POSTGRES_DB:-knowledge_assistant_db}}
      - API_PORT=${API_PORT:-8080}
      # Pass other necessary env vars from .env file
    volumes:
      - ./app:/app/app
    env_file:
      - .env
    command: uvicorn app.api.main:app --host 0.0.0.0 --port ${API_PORT:-8080} --reload

  worker:
    build:
      context: .
      dockerfile: Dockerfile.worker
    depends_on:
      redis:
        condition: service_healthy
      postgres_db:
        condition: service_healthy
      vector_db:
        condition: service_started # Worker directly interacts with vector_db
      api: # Ensure API is at least started if worker needs to call back or for shared code
        condition: service_started
    environment:
      - LOG_LEVEL=${LOG_LEVEL:-INFO}
      - CELERY_BROKER_URL=${CELERY_BROKER_URL:-redis://redis:6379/0}
      - CELERY_RESULT_BACKEND=${CELERY_RESULT_BACKEND:-redis://redis:6379/0}
      - DATABASE_URL=${DATABASE_URL:-postgresql://${POSTGRES_USER:-user}:${POSTGRES_PASSWORD:-password}@postgres_db:5432/${POSTGRES_DB:-knowledge_assistant_db}}
      - VECTOR_DB_IMPL=${VECTOR_DB_IMPL:-chromadb}
      - CHROMA_HOST=${CHROMA_HOST:-vector_db}
      - CHROMA_PORT=${CHROMA_PORT:-8000}
      - CHROMA_COLLECTION_NAME=${CHROMA_COLLECTION_NAME:-simple_wikipedia_chunks}
      - LLM_MODEL_NAME=${LLM_MODEL_NAME:-mistralai/Mistral-7B-Instruct-v0.1}
      - EMBEDDING_MODEL_NAME=${EMBEDDING_MODEL_NAME:-sentence-transformers/all-MiniLM-L6-v2}
      - RETRIEVAL_TOP_K=${RETRIEVAL_TOP_K:-5}
      # Pass other necessary env vars from .env file
      # Mount local models directory if LLM_MODEL_PATH is used
      # - HUGGING_FACE_HUB_TOKEN=${HUGGING_FACE_HUB_TOKEN}
    volumes:
      - ./app:/app/app
      # Example for mounting local models (ensure path exists and matches .env)
      # - ./models_cache:/root/.cache/huggingface/hub # Mount Hugging Face cache
      # - /path/to/your/local/llm_models:/models # If LLM_MODEL_PATH is used
    env_file:
      - .env
    command: watchmedo auto-restart --directory=./app --pattern=*.py --recursive -- celery -A app.worker.celery_app.celery worker -l ${LOG_LEVEL:-INFO} -Q technical_qna --concurrency=1

  # Data ingestion service (run manually or on a schedule)
  ingestion:
    build:
      context: .
      dockerfile: Dockerfile.ingestion # Separate Dockerfile for ingestion if dependencies differ significantly
    depends_on:
      vector_db:
        condition: service_started # Needs vector_db to be up
      postgres_db: # If storing chunks/metadata in postgres
        condition: service_healthy
    environment:
      - LOG_LEVEL=${LOG_LEVEL:-INFO}
      - DATABASE_URL=${DATABASE_URL:-postgresql://${POSTGRES_USER:-user}:${POSTGRES_PASSWORD:-password}@postgres_db:5432/${POSTGRES_DB:-knowledge_assistant_db}}
      - VECTOR_DB_IMPL=${VECTOR_DB_IMPL:-chromadb}
      - CHROMA_HOST=${CHROMA_HOST:-vector_db}
      - CHROMA_PORT=${CHROMA_PORT:-8000}
      - CHROMA_COLLECTION_NAME=${CHROMA_COLLECTION_NAME:-simple_wikipedia_chunks}
      - EMBEDDING_MODEL_NAME=${EMBEDDING_MODEL_NAME:-sentence-transformers/all-MiniLM-L6-v2}
      - DATASET_NAME=${DATASET_NAME:-rahular/simple-wikipedia}
      - CHUNK_SIZE=${CHUNK_SIZE:-512}
      - CHUNK_OVERLAP=${CHUNK_OVERLAP:-64}
      # - HUGGING_FACE_HUB_TOKEN=${HUGGING_FACE_HUB_TOKEN}
    volumes:
      - ./data_ingestion:/app/data_ingestion
      - ./app/core:/app/app/core # Share core config/logging if needed
      # - ./models_cache:/root/.cache/huggingface/hub # Mount Hugging Face cache
    env_file:
      - .env
    # Command to run ingestion, e.g., python data_ingestion/ingest.py
    # This service is intended to be run on demand:
    # docker-compose run --rm ingestion
    command: ["tail", "-f", "/dev/null"] # Keep container running if needed, or specify ingestion command

  prometheus:
    image: prom/prometheus:v2.51.0
    ports:
      - "9090:9090"
    volumes:
      - ./monitoring/prometheus.yml:/etc/prometheus/prometheus.yml
      - 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' # Allows reload of config via HTTP POST to /-/reload
    depends_on:
      - api

  grafana:
    image: grafana/grafana:10.4.2
    ports:
      - "3000:3000"
    volumes:
      - grafana_data:/var/lib/grafana
      - ./monitoring/grafana_datasources.yml:/etc/grafana/provisioning/datasources/default.yml
      # - ./monitoring/grafana_dashboards.yml:/etc/grafana/provisioning/dashboards/default.yml
      # - ./monitoring/dashboards:/var/lib/grafana/dashboards # If you have pre-defined dashboards
    environment:
      GF_SECURITY_ADMIN_USER: ${GRAFANA_ADMIN_USER:-admin}
      GF_SECURITY_ADMIN_PASSWORD: ${GRAFANA_ADMIN_PASSWORD:-password} # Change in production
      GF_DATASOURCES_DEFAULT_VERSION: 1 # Ensures datasource provisioning is picked up
    depends_on:
      - prometheus
    env_file:
      - .env

volumes:
  redis_data:
  postgres_data:
  chroma_data:
  prometheus_data:
  grafana_data:
  # models_cache: # If using a shared Hugging Face cache volume