docker-compose.yml

version: '3.8'

x-json-logging: &json-logging
  logging:
    driver: json-file
    options:
      max-size: "100m"
      max-file: "3"

services:
  # Soroban local network (stellar quickstart)
  soroban:
    image: stellar/quickstart:latest
    command: --local --enable-soroban-rpc
    ports:
      - "8000:8000"
    healthcheck:
      # CMD-SHELL so a failure echoes a greppable WARN line into the health
      # log (docker inspect --format '{{json .State.Health}}').
      test: ["CMD-SHELL", "curl -fsS http://localhost:8000 || (echo 'WARN: soroban health check failed' >&2; exit 1)"]
      interval: 5s
      timeout: 3s
      retries: 30
      # Stellar quickstart takes a while to bring up Horizon + RPC.
      start_period: 30s
    <<: *json-logging

  # MPC Node 0
  mpc-node-0:
    build:
      context: .
      dockerfile: services/node/Dockerfile
    environment:
      NODE_ID: "0"
      PORT: "8101"
      MPC_PORT: "10000"
      PARTY_CONFIG: "/app/config/party_0.toml"
      CIRCUIT_DIR: "/app/circuits"
      CRS_DIR: "/app/crs"
      REQUEST_LOG_FORMAT: "json"
    ports:
      - "8101:8101"
      - "10000:10000"
    volumes:
      - ./crs:/app/crs
      - ./circuits:/app/circuits:ro
      - ./services/node/config:/app/config:ro
    healthcheck:
      test: ["CMD-SHELL", "curl -fsS http://localhost:8101/health || (echo 'WARN: mpc-node-0 health check failed' >&2; exit 1)"]
      interval: 5s
      timeout: 3s
      retries: 30
      # co-noir key generation + CRS load can take 30-60s on a cold start;
      # failures during this window do not count against `retries`.
      start_period: 60s
    <<: *json-logging

  # MPC Node 1
  mpc-node-1:
    build:
      context: .
      dockerfile: services/node/Dockerfile
    environment:
      NODE_ID: "1"
      PORT: "8102"
      MPC_PORT: "10001"
      PARTY_CONFIG: "/app/config/party_1.toml"
      CIRCUIT_DIR: "/app/circuits"
      CRS_DIR: "/app/crs"
      REQUEST_LOG_FORMAT: "json"
    ports:
      - "8102:8102"
      - "10001:10001"
    volumes:
      - ./crs:/app/crs
      - ./circuits:/app/circuits:ro
      - ./services/node/config:/app/config:ro
    healthcheck:
      test: ["CMD-SHELL", "curl -fsS http://localhost:8102/health || (echo 'WARN: mpc-node-1 health check failed' >&2; exit 1)"]
      interval: 5s
      timeout: 3s
      retries: 30
      # co-noir key generation + CRS load can take 30-60s on a cold start;
      # failures during this window do not count against `retries`.
      start_period: 60s
    <<: *json-logging

  # MPC Node 2
  mpc-node-2:
    build:
      context: .
      dockerfile: services/node/Dockerfile
    environment:
      NODE_ID: "2"
      PORT: "8103"
      MPC_PORT: "10002"
      PARTY_CONFIG: "/app/config/party_2.toml"
      CIRCUIT_DIR: "/app/circuits"
      CRS_DIR: "/app/crs"
      REQUEST_LOG_FORMAT: "json"
    ports:
      - "8103:8103"
      - "10002:10002"
    volumes:
      - ./crs:/app/crs
      - ./circuits:/app/circuits:ro
      - ./services/node/config:/app/config:ro
    healthcheck:
      test: ["CMD-SHELL", "curl -fsS http://localhost:8103/health || (echo 'WARN: mpc-node-2 health check failed' >&2; exit 1)"]
      interval: 5s
      timeout: 3s
      retries: 30
      # co-noir key generation + CRS load can take 30-60s on a cold start;
      # failures during this window do not count against `retries`.
      start_period: 60s
    <<: *json-logging

  # Standby MPC nodes (Issue #95 — 5-node committee, N=5 / threshold=3).
  #
  # co-noir's REP3 protocol is fixed at exactly 3 active parties per session
  # (see infrastructure/helm/mpc-node/values.yaml) — that "3" doesn't change.
  # What "N=5, threshold=3" adds is redundancy: 2 extra pre-provisioned nodes
  # that hold the *same* identity/config as an existing role (0 or 1) and can
  # be promoted to take that role's place if its primary container fails,
  # without a re-key ceremony. They are not started by default (`standby`
  # profile) — bring one up with `scripts/promote-mpc-standby.sh <role>`,
  # which stops the failed primary and starts its standby in its place.
  # See docs/mpc-committee-resilience.md.
  mpc-node-3:
    profiles: ["standby"]
    build:
      context: .
      dockerfile: services/node/Dockerfile
    environment:
      NODE_ID: "0" # standby for role 0 — same identity as mpc-node-0
      PORT: "8104"
      MPC_PORT: "10000"
      PARTY_CONFIG: "/app/config/party_0.toml"
      CIRCUIT_DIR: "/app/circuits"
      CRS_DIR: "/app/crs"
      REQUEST_LOG_FORMAT: "json"
    ports:
      - "8104:8104"
    volumes:
      - ./crs:/app/crs
      - ./circuits:/app/circuits:ro
      - ./services/node/config:/app/config:ro
    healthcheck:
      test: ["CMD-SHELL", "curl -fsS http://localhost:8104/health || (echo 'WARN: mpc-node-3 (standby for role 0) health check failed' >&2; exit 1)"]
      interval: 5s
      timeout: 3s
      retries: 30
      start_period: 60s
    <<: *json-logging

  mpc-node-4:
    profiles: ["standby"]
    build:
      context: .
      dockerfile: services/node/Dockerfile
    environment:
      NODE_ID: "1" # standby for role 1 — same identity as mpc-node-1
      PORT: "8105"
      MPC_PORT: "10001"
      PARTY_CONFIG: "/app/config/party_1.toml"
      CIRCUIT_DIR: "/app/circuits"
      CRS_DIR: "/app/crs"
      REQUEST_LOG_FORMAT: "json"
    ports:
      - "8105:8105"
    volumes:
      - ./crs:/app/crs
      - ./circuits:/app/circuits:ro
      - ./services/node/config:/app/config:ro
    healthcheck:
      test: ["CMD-SHELL", "curl -fsS http://localhost:8105/health || (echo 'WARN: mpc-node-4 (standby for role 1) health check failed' >&2; exit 1)"]
      interval: 5s
      timeout: 3s
      retries: 30
      start_period: 60s
    <<: *json-logging

  # Coordinator (orchestrates MPC + serves API to web app)
  coordinator:
    build:
      context: .
      dockerfile: services/coordinator/Dockerfile
    env_file:
      - path: .env.local
        required: false
    environment:
      MPC_NODE_0: "http://mpc-node-0:8101"
      MPC_NODE_1: "http://mpc-node-1:8102"
      MPC_NODE_2: "http://mpc-node-2:8103"
      SOROBAN_RPC: "http://soroban:8000/soroban/rpc"
      CIRCUIT_DIR: "/app/circuits"
      CRS_DIR: "/app/crs"
      BIND_ADDR: "0.0.0.0:8080"
      REQUEST_LOG_FORMAT: "json"
    ports:
      - "8080:8080"
    volumes:
      - ./crs:/app/crs
      - ./circuits:/app/circuits:ro
    depends_on:
      soroban:
        condition: service_healthy
      mpc-node-0:
        condition: service_healthy
      mpc-node-1:
        condition: service_healthy
      mpc-node-2:
        condition: service_healthy
    healthcheck:
      test: ["CMD-SHELL", "curl -fsS http://localhost:8080/api/health || (echo 'WARN: coordinator health check failed' >&2; exit 1)"]
      interval: 5s
      timeout: 3s
      retries: 30
      start_period: 20s
    <<: *json-logging

  # Web app
  # app is deployed on Vercel — run locally with: cd app && npm run dev