docker-compose.e2e.yaml

# MoltNet Docker Compose — E2E Testing
#
# Full stack (infra + apps) with ephemeral storage.
# Data is reset on every restart.
#
# Repo-built app images are NOT built by Compose. They are produced by Nx
# (`docker:build`) and tagged `ghcr.io/getlarge/themoltnet/<svc>:dev` by
# `pnpm run e2e:build`. This file references them via `image:`. See issue #1498
# and AGENTS.md → "Docker image contract". This single file is used both
# locally and in CI; CI overrides the tags by exporting the *_IMAGE env vars
# (e.g. REST_API_IMAGE=ghcr.io/getlarge/themoltnet/rest-api:ci-<sha>).
#
# Usage (local):
#   pnpm run e2e:up      # builds images via Nx, then starts the stack
#   pnpm run e2e:down    # stops + removes volumes
#   pnpm run e2e:reset   # down + up
#
# Or directly, after `pnpm run e2e:build`:
#   COMPOSE_DISABLE_ENV_FILE=true docker compose -f docker-compose.e2e.yaml up -d
#   COMPOSE_DISABLE_ENV_FILE=true docker compose -f docker-compose.e2e.yaml down -v

services:
  app-db:
    extends:
      file: docker-compose.base.yaml
      service: app-db

  app-db-migrate:
    extends:
      file: docker-compose.base.yaml
      service: app-db-migrate
    # Use the Nx-built image (db-migrate) instead of Compose building the base
    # service's Dockerfile. !reset null drops the inherited build: block.
    build: !reset null
    image: ${DB_MIGRATE_IMAGE:-ghcr.io/getlarge/themoltnet/db-migrate:dev}
    depends_on:
      app-db:
        condition: service_healthy
    restart: 'on-failure:3'

  issue-lifecycle-db:
    extends:
      file: docker-compose.base.yaml
      service: issue-lifecycle-db

  issue-lifecycle-db-migrate:
    extends:
      file: docker-compose.base.yaml
      service: issue-lifecycle-db-migrate
    depends_on:
      issue-lifecycle-db:
        condition: service_healthy
    restart: 'on-failure:3'

  kratos-postgres:
    extends:
      file: docker-compose.base.yaml
      service: kratos-postgres

  kratos-migrate:
    extends:
      file: docker-compose.base.yaml
      service: kratos-migrate
    depends_on:
      kratos-postgres:
        condition: service_healthy
    restart: 'on-failure:3'

  kratos:
    extends:
      file: docker-compose.base.yaml
      service: kratos
    environment:
      SELFSERVICE_FLOWS_VERIFICATION_ENABLED: 'true'
      SELFSERVICE_FLOWS_REGISTRATION_AFTER_OIDC_HOOKS_0_CONFIG_URL: http://rest-api:8080/hooks/kratos/after-registration
      SELFSERVICE_FLOWS_REGISTRATION_AFTER_PASSWORD_HOOKS_0_CONFIG_URL: http://rest-api:8080/hooks/kratos/after-registration
      SELFSERVICE_FLOWS_LOGIN_AFTER_OIDC_HOOKS_0_CONFIG_URL: http://rest-api:8080/hooks/kratos/after-login
      SELFSERVICE_FLOWS_LOGIN_AFTER_PASSWORD_HOOKS_0_CONFIG_URL: http://rest-api:8080/hooks/kratos/after-login
      SELFSERVICE_FLOWS_SETTINGS_AFTER_PASSWORD_HOOKS_0_CONFIG_URL: http://rest-api:8080/hooks/kratos/validate-settings
      SELFSERVICE_FLOWS_SETTINGS_AFTER_PASSWORD_HOOKS_1_CONFIG_URL: http://rest-api:8080/hooks/kratos/after-settings
      SELFSERVICE_FLOWS_SETTINGS_AFTER_PROFILE_HOOKS_0_CONFIG_URL: http://rest-api:8080/hooks/kratos/validate-settings
      SELFSERVICE_FLOWS_SETTINGS_AFTER_PROFILE_HOOKS_1_CONFIG_URL: http://rest-api:8080/hooks/kratos/after-settings
    depends_on:
      kratos-migrate:
        condition: service_completed_successfully
      mailslurper:
        condition: service_started

  kratos-selfservice-ui-node:
    extends:
      file: docker-compose.base.yaml
      service: kratos-selfservice-ui-node
    environment:
      KRATOS_PUBLIC_URL: http://kratos:4433/
    depends_on:
      kratos:
        condition: service_healthy

  hydra-postgres:
    extends:
      file: docker-compose.base.yaml
      service: hydra-postgres

  hydra-migrate:
    extends:
      file: docker-compose.base.yaml
      service: hydra-migrate
    depends_on:
      hydra-postgres:
        condition: service_healthy
    restart: 'on-failure:3'

  hydra:
    extends:
      file: docker-compose.base.yaml
      service: hydra
    environment:
      URLS_SELF_ISSUER: http://hydra:4444
      URLS_SELF_PUBLIC: http://hydra:4444
    depends_on:
      hydra-migrate:
        condition: service_completed_successfully

  keto-postgres:
    extends:
      file: docker-compose.base.yaml
      service: keto-postgres

  keto-migrate:
    extends:
      file: docker-compose.base.yaml
      service: keto-migrate
    depends_on:
      keto-postgres:
        condition: service_healthy
    restart: 'on-failure:3'

  keto:
    extends:
      file: docker-compose.base.yaml
      service: keto
    depends_on:
      keto-migrate:
        condition: service_completed_successfully

  # Talos declares /var/lib/talos as an image VOLUME. Override it with tmpfs so
  # CI/e2e cannot reuse a development named volume and all state is ephemeral.
  talos:
    extends:
      file: docker-compose.base.yaml
      service: talos
    tmpfs:
      - /var/lib/talos:uid=65532,gid=65532,mode=0700

  mailslurper:
    extends:
      file: docker-compose.base.yaml
      service: mailslurper

  redis:
    extends:
      file: docker-compose.base.yaml
      service: redis
    # Host port (6380 to avoid colliding with a local dev redis on 6379) so the
    # rate-limit e2e can inspect that limiter keys actually land in Redis.
    ports:
      - '6380:6379'

  runtime-object-store:
    extends:
      file: docker-compose.base.yaml
      service: runtime-object-store

  rest-api:
    image: ${REST_API_IMAGE:-ghcr.io/getlarge/themoltnet/rest-api:dev}
    ports:
      - '8080:8080'
    environment:
      NODE_ENV: test
      PORT: '8080'
      DATABASE_URL: postgresql://moltnet:moltnet_secret@app-db:5432/moltnet
      DBOS_SYSTEM_DATABASE_URL: postgresql://moltnet:moltnet_secret@app-db:5432/moltnet?schema=dbos
      ORY_KRATOS_PUBLIC_URL: http://kratos:4433
      ORY_KRATOS_ADMIN_URL: http://kratos:4434
      ORY_HYDRA_PUBLIC_URL: http://hydra:4444
      ORY_HYDRA_ADMIN_URL: http://hydra:4445
      ORY_KETO_PUBLIC_URL: http://keto:4466
      ORY_KETO_ADMIN_URL: http://keto:4467
      ORY_TALOS_ADMIN_URL: http://talos:4420
      # Integration tests assert immediate revocation. Production keeps the
      # documented 60-second positive-cache window; TTL zero still exercises
      # single-flight request coalescing without retaining positive entries.
      ORY_AUTH_CACHE_TTL_MS: '0'
      ORY_ACTION_API_KEY: local-dev-webhook-key
      RECOVERY_CHALLENGE_SECRET: e2e-recovery-secret-for-hmac-signing
      CORS_ORIGINS: 'http://localhost:5174,http://localhost:4433'
      AUTH_SCOPE_ENFORCEMENT: 'enforce'
      RATE_LIMIT_GLOBAL_AUTH: '10000'
      RATE_LIMIT_GLOBAL_ANON: '10000'
      # Distinct from GLOBAL_AUTH so the read/write-split e2e can assert the read
      # bucket's limit header differs from the mutation bucket's. High enough that
      # read-heavy sibling suites never trip it (shared per-identity read bucket).
      RATE_LIMIT_GLOBAL_READ: '7000'
      RATE_LIMIT_EMBEDDING: '10000'
      RATE_LIMIT_VOUCH: '10000'
      RATE_LIMIT_SIGNING: '10000'
      SIGNING_MAX_PENDING_REQUESTS: '10'
      RATE_LIMIT_AGENT_KEY: '10000'
      RATE_LIMIT_TASK_ARTIFACT_UPLOAD: '10000'
      RATE_LIMIT_RECOVERY: '10000'
      # High enough for the full suite's legitimate verification traffic. The
      # adversarial rate-limit e2e seeds its isolated Redis counter immediately
      # below this limit, proving a real shared-bucket 429 without poisoning
      # sibling Ed25519 and previewSign tests.
      RATE_LIMIT_PUBLIC_VERIFY: '10000'
      RATE_LIMIT_PUBLIC_SEARCH: '10000'
      RATE_LIMIT_LEGREFFIER_START: '10000'
      RATE_LIMIT_REGISTRATION: '10000'
      # Keep the pre-resolve IP throttle high: every suite shares the localhost
      # IP, so a low value here would throttle unrelated suites mid-run.
      RATE_LIMIT_PRE_RESOLVE_IP: '10000'
      # Injected by e2e globalSetup after bootstrapping a sponsor genesis agent.
      # Empty by default — server starts cleanly (503 on /public/legreffier/start)
      # until globalSetup restarts with SPONSOR_AGENT_ID set.
      SPONSOR_AGENT_ID: ${SPONSOR_AGENT_ID:-}
      # Back the main rate limiter with the shared Redis store (Part 3). Exercises
      # the Redis-backed path in e2e; without these the limiter is in-memory.
      REDIS_HOST: redis
      REDIS_PORT: '6379'
      RUNTIME_SESSION_STORAGE_ENDPOINT: http://runtime-object-store:8333
      RUNTIME_SESSION_STORAGE_REGION: auto
      RUNTIME_SESSION_STORAGE_BUCKET: moltnet-runtime-sessions-e2e
      RUNTIME_SESSION_STORAGE_ACCESS_KEY_ID: runtime-sessions
      RUNTIME_SESSION_STORAGE_SECRET_ACCESS_KEY: runtime-sessions-secret
      RUNTIME_SESSION_STORAGE_FORCE_PATH_STYLE: 'true'
      RUNTIME_SESSION_MAX_BYTES: '10485760'
      TASK_ARTIFACT_STORAGE_ENDPOINT: http://runtime-object-store:8333
      TASK_ARTIFACT_STORAGE_REGION: auto
      TASK_ARTIFACT_STORAGE_BUCKET: moltnet-task-artifacts-e2e
      TASK_ARTIFACT_STORAGE_ACCESS_KEY_ID: runtime-sessions
      TASK_ARTIFACT_STORAGE_SECRET_ACCESS_KEY: runtime-sessions-secret
      TASK_ARTIFACT_STORAGE_FORCE_PATH_STYLE: 'true'
      TASK_ARTIFACT_MAX_BYTES: '26214400'
    depends_on:
      app-db:
        condition: service_healthy
      app-db-migrate:
        condition: service_completed_successfully
      kratos:
        condition: service_healthy
      hydra:
        condition: service_healthy
      keto:
        condition: service_healthy
      talos:
        # Talos auth is optional; REST startup must not block unrelated suites.
        # Talos-specific e2e setup waits by calling its admin API directly.
        condition: service_started
      redis:
        condition: service_healthy
      runtime-object-store:
        condition: service_started
    healthcheck:
      test:
        [
          'CMD',
          'node',
          '-e',
          "fetch('http://localhost:8080/health').then(r=>r.ok?process.exit(0):process.exit(1)).catch(()=>process.exit(1))",
        ]
      interval: 5s
      timeout: 3s
      retries: 15
      start_period: 30s

  console:
    image: ${CONSOLE_IMAGE:-ghcr.io/getlarge/themoltnet/console:dev}
    ports:
      - '5174:80'
    environment:
      KRATOS_PUBLIC_URL: http://localhost:4433
      API_BASE_URL: http://localhost:8080
      CONSOLE_BASE_URL: http://localhost:5174
      SIGNER_BASE_URL: http://127.0.0.1:17373
      PACK_GC_COMPILE_TTL_DAYS: '7'
    depends_on:
      rest-api:
        condition: service_healthy
      kratos:
        condition: service_healthy

  mcp-server:
    image: ${MCP_SERVER_IMAGE:-ghcr.io/getlarge/themoltnet/mcp-server:dev}
    ports:
      - '8001:8001'
    environment:
      NODE_ENV: production
      PORT: '8001'
      REST_API_URL: http://rest-api:8080
      CONSOLE_BASE_URL: http://localhost:5174
      AUTH_ENABLED: 'true'
      ORY_PROJECT_URL: http://hydra:4444
      CLIENT_CREDENTIALS_PROXY: 'true'
      MCP_REDIS_HOST: redis
      MCP_REDIS_PORT: '6379'
      MCP_APP_DOMAIN: http://localhost:8001
      MCP_APP_CONNECT_DOMAINS: http://localhost:8001,http://rest-api:8080
    depends_on:
      rest-api:
        condition: service_healthy
      hydra:
        condition: service_healthy
      redis:
        condition: service_healthy
    healthcheck:
      test:
        [
          'CMD',
          'node',
          '-e',
          "fetch('http://localhost:8001/healthz').then(r=>r.ok?process.exit(0):process.exit(1)).catch(()=>process.exit(1))",
        ]
      interval: 5s
      timeout: 3s
      retries: 15
      start_period: 15s

  mcp-host:
    image: ${MCP_HOST_IMAGE:-ghcr.io/getlarge/themoltnet/mcp-host:dev}
    ports:
      - '8082:8080'
      - '8083:8081'
    environment:
      AUTORUN: '0'
      DEFAULT_ARGS: '{}'
      DEFAULT_TOOL: tasks_app_open
      MCP_SERVER_URL: http://localhost:8001/mcp
      SANDBOX_BASE_URL: http://localhost:8083/sandbox.html
    depends_on:
      mcp-server:
        condition: service_healthy
    healthcheck:
      test:
        [
          'CMD',
          'node',
          '-e',
          "fetch('http://localhost:8080/healthz').then(r=>r.ok?process.exit(0):process.exit(1)).catch(()=>process.exit(1))",
        ]
      interval: 5s
      timeout: 3s
      retries: 15
      start_period: 15s