docker-compose.yml

# =============================================================================
# ARIA — Autonomous Research & Intelligence Assistant
# H:\Training\AI_Engineering\ARIA\docker-compose.yml
#
# Services
# ─────────────────────────────────────────────────────────────────────────────
# Inherited from Langfuse_POC:
#   langfuse-web     :3000   LLM observability UI + API
#   langfuse-worker  :3030   Async trace processor
#   postgres         :5432   Langfuse metadata + MLflow backend
#   clickhouse       :8123   OLAP trace store (Langfuse v3)
#   minio            :9090   S3-compatible blob store
#   mlflow           :5000   Experiment tracking
#   ollama           :11434  Local LLM runtime
#   open-webui       :8080   Chat UI (Ollama + OpenRouter)
#   litellm          :4000   LLM proxy / router
#
# New for ARIA:
#   redis            :6379   Session state + A2A message bus
#   qdrant           :6333   Vector store (RAG + long-term memory)
#   aria-api         :8000   FastAPI orchestration entrypoint
#   pipelines        :9099   OpenWebUI Pipelines (ARIA chat model)\n#   pipelines        :9099   OpenWebUI Pipelines (ARIA chat model)
# =============================================================================

name: aria

# ── Shared env anchors ────────────────────────────────────────────────────────
x-langfuse-env: &langfuse-env
  DATABASE_URL: postgresql://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:-postgres}@postgres:5432/${POSTGRES_DB:-langfuse}
  NEXTAUTH_URL: ${NEXTAUTH_URL:-http://localhost:3000}
  NEXTAUTH_SECRET: ${NEXTAUTH_SECRET:-changeme-generate-with-openssl-rand-base64-32}
  SALT: ${SALT:-changeme-generate-with-openssl-rand-base64-16}
  ENCRYPTION_KEY: ${ENCRYPTION_KEY:-0000000000000000000000000000000000000000000000000000000000000000}
  TELEMETRY_ENABLED: "false"
  LANGFUSE_ENABLE_EXPERIMENTAL_FEATURES: "true"
  # ClickHouse
  CLICKHOUSE_MIGRATION_URL: clickhouse://clickhouse:9000
  CLICKHOUSE_URL: http://clickhouse:8123
  CLICKHOUSE_USER: ${CLICKHOUSE_USER:-clickhouse}
  CLICKHOUSE_PASSWORD: ${CLICKHOUSE_PASSWORD:-clickhouse}
  CLICKHOUSE_CLUSTER_ENABLED: "false"
  # Redis (shared with ARIA)
  REDIS_HOST: redis
  REDIS_PORT: "6379"
  REDIS_AUTH: ${REDIS_PASSWORD:-}
  REDIS_TLS_ENABLED: "false"
  # MinIO / S3
  LANGFUSE_S3_EVENT_UPLOAD_BUCKET: langfuse
  LANGFUSE_S3_EVENT_UPLOAD_REGION: auto
  LANGFUSE_S3_EVENT_UPLOAD_ACCESS_KEY_ID: ${MINIO_ROOT_USER:-minio}
  LANGFUSE_S3_EVENT_UPLOAD_SECRET_ACCESS_KEY: ${MINIO_ROOT_PASSWORD:-miniosecret}
  LANGFUSE_S3_EVENT_UPLOAD_ENDPOINT: http://minio:9000
  LANGFUSE_S3_EVENT_UPLOAD_FORCE_PATH_STYLE: "true"
  LANGFUSE_S3_MEDIA_UPLOAD_BUCKET: langfuse
  LANGFUSE_S3_MEDIA_UPLOAD_REGION: auto
  LANGFUSE_S3_MEDIA_UPLOAD_ACCESS_KEY_ID: ${MINIO_ROOT_USER:-minio}
  LANGFUSE_S3_MEDIA_UPLOAD_SECRET_ACCESS_KEY: ${MINIO_ROOT_PASSWORD:-miniosecret}
  LANGFUSE_S3_MEDIA_UPLOAD_ENDPOINT: http://minio:9000
  LANGFUSE_S3_MEDIA_UPLOAD_FORCE_PATH_STYLE: "true"
  LANGFUSE_S3_BATCH_EXPORT_BUCKET: langfuse
  LANGFUSE_S3_BATCH_EXPORT_REGION: auto
  LANGFUSE_S3_BATCH_EXPORT_ACCESS_KEY_ID: ${MINIO_ROOT_USER:-minio}
  LANGFUSE_S3_BATCH_EXPORT_SECRET_ACCESS_KEY: ${MINIO_ROOT_PASSWORD:-miniosecret}
  LANGFUSE_S3_BATCH_EXPORT_ENDPOINT: http://minio:9000
  LANGFUSE_S3_BATCH_EXPORT_FORCE_PATH_STYLE: "true"
  # Auto-init org/project (no manual UI setup needed on first boot)
  LANGFUSE_INIT_ORG_ID: aria-org
  LANGFUSE_INIT_ORG_NAME: ARIA
  LANGFUSE_INIT_PROJECT_ID: aria-project
  LANGFUSE_INIT_PROJECT_NAME: aria
  LANGFUSE_INIT_PROJECT_PUBLIC_KEY: ${LANGFUSE_PUBLIC_KEY:-pk-lf-local}
  LANGFUSE_INIT_PROJECT_SECRET_KEY: ${LANGFUSE_SECRET_KEY:-sk-lf-local}
  LANGFUSE_INIT_USER_EMAIL: ${LANGFUSE_INIT_USER_EMAIL:-admin@local.dev}
  LANGFUSE_INIT_USER_PASSWORD: ${LANGFUSE_INIT_USER_PASSWORD:-admin}
  LANGFUSE_INIT_USER_NAME: ${LANGFUSE_INIT_USER_NAME:-Admin}

services:

  # ── INFRASTRUCTURE ──────────────────────────────────────────────────────────

  postgres:
    image: postgres:17-alpine
    container_name: aria-postgres
    restart: unless-stopped
    environment:
      POSTGRES_USER: ${POSTGRES_USER:-postgres}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-postgres}
      POSTGRES_DB: ${POSTGRES_DB:-langfuse}
    volumes:
      - postgres-data:/var/lib/postgresql/data
      # Init script: create mlflow DB alongside langfuse DB
      - ./init-db:/docker-entrypoint-initdb.d:ro
    ports:
      - "127.0.0.1:5432:5432"
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-postgres} -d ${POSTGRES_DB:-langfuse}"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - aria-net

  clickhouse:
    image: clickhouse/clickhouse-server:24.12-alpine
    container_name: aria-clickhouse
    restart: unless-stopped
    environment:
      CLICKHOUSE_DB: default
      CLICKHOUSE_USER: ${CLICKHOUSE_USER:-clickhouse}
      CLICKHOUSE_PASSWORD: ${CLICKHOUSE_PASSWORD:-clickhouse}
    volumes:
      - clickhouse-data:/var/lib/clickhouse
      - clickhouse-logs:/var/log/clickhouse-server
    ports:
      - "127.0.0.1:18123:8123"
      - "127.0.0.1:19000:9000"
    healthcheck:
      test: ["CMD-SHELL", "clickhouse-client --user ${CLICKHOUSE_USER:-clickhouse} --password ${CLICKHOUSE_PASSWORD:-clickhouse} --query 'SELECT 1'"]
      interval: 10s
      timeout: 5s
      retries: 10
      start_period: 60s
    ulimits:
      nofile:
        soft: 262144
        hard: 262144
    networks:
      - aria-net

  minio:
    image: minio/minio:latest
    container_name: aria-minio
    restart: unless-stopped
    command: server /data --console-address ":9091"
    environment:
      MINIO_ROOT_USER: ${MINIO_ROOT_USER:-minio}
      MINIO_ROOT_PASSWORD: ${MINIO_ROOT_PASSWORD:-miniosecret}
    volumes:
      - minio-data:/data
    ports:
      - "9090:9000"   # S3 API
      - "9091:9091"   # MinIO console UI
    healthcheck:
      test: ["CMD", "mc", "ready", "local"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - aria-net

  # MinIO bucket init (runs once, exits)
  minio-init:
    image: minio/mc:latest
    container_name: aria-minio-init
    depends_on:
      minio:
        condition: service_healthy
    entrypoint: >
      /bin/sh -c "
        mc alias set local http://minio:9000 ${MINIO_ROOT_USER:-minio} ${MINIO_ROOT_PASSWORD:-miniosecret};
        mc mb --ignore-existing local/langfuse;
        mc mb --ignore-existing local/mlflow;
        exit 0
      "
    networks:
      - aria-net

  # Redis — shared by Langfuse v3, ARIA session state, and A2A message bus
  redis:
    image: redis:7-alpine
    container_name: aria-redis
    restart: unless-stopped
    command: >
      redis-server
      --appendonly yes
      --maxmemory 512mb
      --maxmemory-policy allkeys-lru
      ${REDIS_PASSWORD:+--requirepass ${REDIS_PASSWORD}}
    volumes:
      - redis-data:/data
    ports:
      - "127.0.0.1:6379:6379"
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 5s
      timeout: 3s
      retries: 5
    networks:
      - aria-net

  # Qdrant — vector store for RAG and long-term agent memory
  qdrant:
    image: qdrant/qdrant:latest
    container_name: aria-qdrant
    restart: unless-stopped
    volumes:
      - qdrant-data:/qdrant/storage
    ports:
      - "6333:6333"   # REST API + dashboard  → http://localhost:6333/dashboard
      - "127.0.0.1:6334:6334"   # gRPC
    healthcheck:
      test: ["CMD-SHELL", "bash -c 'exec 3<>/dev/tcp/localhost/6333 && echo -e \"GET /readyz HTTP/1.1\\r\\nHost: localhost\\r\\nConnection: close\\r\\n\\r\\n\" >&3 && grep -q \"200 OK\" <&3' || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - aria-net

  # ── LLM OBSERVABILITY (Langfuse v3) ─────────────────────────────────────────

  langfuse-web:
    image: langfuse/langfuse:3
    container_name: aria-langfuse-web
    restart: unless-stopped
    depends_on:
      postgres:
        condition: service_healthy
      minio:
        condition: service_healthy
      redis:
        condition: service_healthy
      clickhouse:
        condition: service_healthy
    environment:
      <<: *langfuse-env
    ports:
      - "3000:3000"   # Langfuse UI → http://localhost:3000
    networks:
      - aria-net

  langfuse-worker:
    image: langfuse/langfuse-worker:3
    container_name: aria-langfuse-worker
    restart: unless-stopped
    depends_on:
      postgres:
        condition: service_healthy
      minio:
        condition: service_healthy
      redis:
        condition: service_healthy
      clickhouse:
        condition: service_healthy
    environment:
      <<: *langfuse-env
    ports:
      - "127.0.0.1:3030:3030"
    networks:
      - aria-net

  # ── EXPERIMENT TRACKING (MLflow) ─────────────────────────────────────────────

  mlflow:
    image: ghcr.io/mlflow/mlflow:v2.15.1
    container_name: aria-mlflow
    restart: unless-stopped
    depends_on:
      postgres:
        condition: service_healthy
      minio-init:
        condition: service_completed_successfully
    entrypoint: ["/bin/sh", "-c"]
    command:
      - >
        pip install --no-cache-dir psycopg2-binary &&
        mlflow server
        --backend-store-uri postgresql://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:-postgres}@postgres:5432/mlflow
        --artifacts-destination s3://mlflow
        --host 0.0.0.0
        --port 5000
    environment:
      MLFLOW_S3_ENDPOINT_URL: http://minio:9000
      AWS_ACCESS_KEY_ID: ${MINIO_ROOT_USER:-minio}
      AWS_SECRET_ACCESS_KEY: ${MINIO_ROOT_PASSWORD:-miniosecret}
    ports:
      - "5000:5000"   # MLflow UI → http://localhost:5000
    networks:
      - aria-net

  # ── LOCAL LLM RUNTIME (Ollama) ───────────────────────────────────────────────

  ollama:
    image: ollama/ollama:latest
    container_name: aria-ollama
    restart: unless-stopped
    volumes:
      - ollama-data:/root/.ollama
    ports:
      - "11434:11434"
    # Remove the deploy block if you don't have an NVIDIA GPU
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    healthcheck:
      test: ["CMD-SHELL", "ollama list || exit 1"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 30s
    networks:
      - aria-net

  # ── LLM PROXY (LiteLLM) ──────────────────────────────────────────────────────
  # Routes all agent LLM calls: OpenRouter (remote) + Ollama (local fallback)
  # Manages keys, rate limits, cost tracking, and model aliases

  litellm:
    image: ghcr.io/berriai/litellm:main-latest
    container_name: aria-litellm
    restart: unless-stopped
    depends_on:
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
      ollama:
        condition: service_healthy
    volumes:
      - ./litellm_config.yaml:/app/config.yaml:ro
    command: --config /app/config.yaml --port 4000 --detailed_debug
    environment:
      OPENROUTER_API_KEY: ${OPENROUTER_API_KEY:-}
      LITELLM_MASTER_KEY: ${LITELLM_MASTER_KEY:-sk-aria-master}
      DATABASE_URL: postgresql://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:-postgres}@postgres:5432/litellm
      REDIS_HOST: redis
      REDIS_PORT: "6379"
      REDIS_PASSWORD: ${REDIS_PASSWORD:-}
      # Forward traces to Langfuse
      LANGFUSE_PUBLIC_KEY: ${LANGFUSE_PUBLIC_KEY:-pk-lf-local}
      LANGFUSE_SECRET_KEY: ${LANGFUSE_SECRET_KEY:-sk-lf-local}
      LANGFUSE_HOST: http://langfuse-web:3000
    ports:
      - "4000:4000"   # LiteLLM proxy → http://localhost:4000
    healthcheck:
      test: ["CMD-SHELL", "python -c \"import urllib.request; urllib.request.urlopen('http://localhost:4000/health/liveliness')\" || exit 1"]
      interval: 15s
      timeout: 10s
      retries: 5
      start_period: 20s
    networks:
      - aria-net

  # ── CHAT UI (Open WebUI) ──────────────────────────────────────────────────────
  # Talks to LiteLLM proxy (not Ollama directly) so all calls are traced

  # ── ARIA PIPELINES ────────────────────────────────────────────────────────────
  # OpenWebUI Pipelines server — exposes ARIA as a chat model in OpenWebUI.
  # After startup, add http://pipelines:9099 in OpenWebUI Admin → Connections.

  pipelines:
    image: ghcr.io/open-webui/pipelines:main
    container_name: aria-pipelines
    restart: unless-stopped
    volumes:
      - ./pipelines:/app/pipelines
    environment:
      PIPELINES_DIR: /app/pipelines
      ARIA_BASE_URL: http://aria-api:8000
    ports:
      - "9099:9099"   # Pipelines API → http://localhost:9099
    networks:
      - aria-net

  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: aria-open-webui
    restart: unless-stopped
    depends_on:
      litellm:
        condition: service_healthy
      pipelines:
        condition: service_started
    volumes:
      - open-webui-data:/app/backend/data
    environment:
      # Point at LiteLLM so OpenRouter models appear alongside Ollama
      OPENAI_API_BASE_URL: http://litellm:4000/v1
      OPENAI_API_KEY: ${LITELLM_MASTER_KEY:-sk-aria-master}
      # Also expose raw Ollama for model management
      OLLAMA_BASE_URL: http://ollama:11434
      WEBUI_SECRET_KEY: ${WEBUI_SECRET_KEY:-changeme}
    ports:
      - "8080:8080"   # Open WebUI → http://localhost:8080
    networks:
      - aria-net

  # ── ARIA ORCHESTRATION API ────────────────────────────────────────────────────
  # FastAPI app: POST /run  GET /status/{id}  POST /resume/{id}  GET /sessions

  aria-api:
    build:
      context: .
      dockerfile: Dockerfile
    container_name: aria-api
    restart: unless-stopped
    depends_on:
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
      qdrant:
        condition: service_healthy
      litellm:
        condition: service_healthy
      langfuse-web:
        condition: service_started
    volumes:
      - .:/app
    environment:
      # LLM (via LiteLLM proxy inside the stack)
      OPENAI_API_BASE: http://litellm:4000/v1
      OPENAI_API_KEY: ${LITELLM_MASTER_KEY:-sk-aria-master}
      # Tools
      TAVILY_API_KEY: ${TAVILY_API_KEY:-}
      E2B_API_KEY: ${E2B_API_KEY:-}
      # Memory
      REDIS_URL: redis://${REDIS_PASSWORD:+:${REDIS_PASSWORD}@}redis:6379/0
      QDRANT_URL: http://qdrant:6333
      QDRANT_COLLECTION: ${QDRANT_COLLECTION:-aria_memory}
      # Observability
      LANGFUSE_PUBLIC_KEY: ${LANGFUSE_PUBLIC_KEY:-pk-lf-local}
      LANGFUSE_SECRET_KEY: ${LANGFUSE_SECRET_KEY:-sk-lf-local}
      LANGFUSE_HOST: http://langfuse-web:3000
      # MLflow
      MLFLOW_TRACKING_URI: http://mlflow:5000
      # App
      LOG_LEVEL: ${LOG_LEVEL:-INFO}
      ARIA_DEFAULT_MODEL: ${ARIA_DEFAULT_MODEL:-anthropic/claude-sonnet-4-6}
    ports:
      - "8000:8000"   # ARIA API → http://localhost:8000/docs
    healthcheck:
      test: ["CMD-SHELL", "curl -sf http://localhost:8000/health || exit 1"]
      interval: 15s
      timeout: 5s
      retries: 5
      start_period: 30s
    networks:
      - aria-net

# ── VOLUMES ────────────────────────────────────────────────────────────────────

volumes:
  postgres-data:
  clickhouse-data:
  clickhouse-logs:
  minio-data:
  redis-data:
  qdrant-data:
  ollama-data:
  open-webui-data:

# ── NETWORK ────────────────────────────────────────────────────────────────────

networks:
  aria-net:
    driver: bridge