chore(infra): docker dev-apps stack + apollo-router v1.45 compat

- 新增 docker-compose.dev-apps.yml：7 个应用服务容器化
  （iam/classes/core-edu/content/msg/push-gateway/api-gateway）
  + apollo-router override
- 新增 apollo-router dev-apps-entrypoint.sh / dev-apps-supergraph.yaml
  仅等待 5 个子图（iam/core-edu/content/msg/config-service）
  避免等待未启动的 ai/data-ana
- router.yaml：移除 v1.45 不兼容键
  （require_manifest/manifest_path/max_cost/max_batch_size）
  移除 ELv2 受限特性（persisted_queries/limits.max_depth）
  自托管无 GraphOS license
- docker-compose.yml：config-service 添加 NODE_ENV=production
  修复 pino-pretty devDep 误用
  apollo-router 添加 APOLLO_PERSISTED_QUERIES env
- services/{classes,core-edu,content,config-service}/Dockerfile
  重写为 repo 根 context + shared-ts/shared-proto 工作区依赖
  + --ignore-scripts
  修复 ERR_PNPM_WORKSPACE_PKG_NOT_FOUND 与 ERR_PNPM_IGNORED_BUILDS
- api-gateway healthcheck：wget --spider 改为 wget -O /dev/null
  修复 busybox 对 200+body 误报 exit 8

infra/apollo-router/Dockerfile

@@ -26,13 +26,14 @@ RUN curl -sSL -o /tmp/rover.tar.gz https://github.com/apollographql/rover/releas
 # 预下载 supergraph 插件（federation-rs supergraph v2.9.0）
 # 避免运行时因网络问题（如 GitHub 不可达）导致 rover compose 失败
 # 插件会被缓存到 /root/.rover/bin/supergraph-v2.9.0
+# tarball 内文件路径为 dist/supergraph，需 --strip-components=1 解包
 RUN APOLLO_ELV2_LICENSE=accept rover supergraph compose --config /dev/null --output /tmp/test-supergraph.graphql 2>&1 || \
     (echo "[build] Pre-downloading supergraph plugin..." && \
      curl -sSL -o /tmp/supergraph-plugin.tar.gz \
        "https://github.com/apollographql/federation-rs/releases/download/supergraph%40v2.9.0/supergraph-v2.9.0-x86_64-unknown-linux-gnu.tar.gz" && \
      mkdir -p /root/.rover/bin && \
-     tar -xzf /tmp/supergraph-plugin.tar.gz -C /root/.rover/bin/ && \
+     tar -xzf /tmp/supergraph-plugin.tar.gz --strip-components=1 -C /root/.rover/bin/ dist/supergraph && \
-     mv /root/.rover/bin/supergraph /root/.rover/bin/supergraph-v2.9.0 2>/dev/null || true && \
+     mv /root/.rover/bin/supergraph /root/.rover/bin/supergraph-v2.9.0 && \
      chmod +x /root/.rover/bin/supergraph-v2.9.0 && \
      rm -f /tmp/supergraph-plugin.tar.gz /tmp/test-supergraph.graphql && \
      echo "[build] Supergraph plugin pre-installed")

infra/apollo-router/dev-apps-entrypoint.sh

@@ -0,0 +1,92 @@
+#!/bin/bash
+# Apollo Router 启动脚本（dev-apps 模式：仅 5 个 TS 子图）
+#
+# 与 entrypoint.sh 的差异：
+#   - SUBGRAPHS 仅包含 iam/core-edu/content/msg/config-service（排除 ai/data-ana）
+#   - 依赖 docker compose depends_on 已确保子图健康，等待重试次数减少到 30
+#
+# 使用：通过 docker-compose.dev-apps.yml 卷挂载覆盖 /dist/entrypoint.sh
+
+set -e
+
+SUBGRAPHS=(
+  "iam:iam:3002"
+  "core-edu:core-edu:3004"
+  "content:content:3005"
+  "msg:msg:3007"
+  "config-service:config-service:3011"
+)
+
+echo "[apollo-router:dev-apps] Waiting for subgraphs to be ready..."
+
+wait_for_subgraph() {
+  local name="$1"
+  local host="$2"
+  local port="$3"
+  local max_retries=30
+  local retry=0
+
+  while [ $retry -lt $max_retries ]; do
+    if curl -sf -o /dev/null "http://${host}:${port}/healthz" 2>/dev/null; then
+      echo "[apollo-router:dev-apps] ${name} is ready (http://${host}:${port})"
+      return 0
+    fi
+    retry=$((retry + 1))
+    echo "[apollo-router:dev-apps] Waiting for ${name} at ${host}:${port} (attempt ${retry}/${max_retries})"
+    sleep 3
+  done
+
+  echo "[apollo-router:dev-apps] WARNING: ${name} not ready after ${max_retries} attempts, continuing anyway..."
+  return 0
+}
+
+for entry in "${SUBGRAPHS[@]}"; do
+  name="${entry%%:*}"
+  hostport="${entry#*:}"
+  host="${hostport%%:*}"
+  port="${hostport##*:}"
+  wait_for_subgraph "$name" "$host" "$port"
+done
+
+# 组合 supergraph SDL
+echo "[apollo-router:dev-apps] Composing supergraph SDL..."
+export APOLLO_ELV2_LICENSE=accept
+
+max_compose_retries=5
+compose_retry=0
+while [ $compose_retry -lt $max_compose_retries ]; do
+  compose_retry=$((compose_retry + 1))
+  if rover supergraph compose --config /dist/supergraph.yaml --output /tmp/supergraph.graphql 2>&1; then
+    echo "[apollo-router:dev-apps] Supergraph composed successfully"
+    break
+  fi
+  echo "[apollo-router:dev-apps] Compose attempt ${compose_retry}/${max_compose_retries} failed, retrying in 5s..."
+  sleep 5
+done
+
+if [ ! -f /tmp/supergraph.graphql ]; then
+  echo "[apollo-router:dev-apps] ERROR: Failed to compose supergraph after ${max_compose_retries} attempts"
+  exit 1
+fi
+
+# PQ manifest 检查（dev 模式 require_manifest=false，仅警告）
+MANIFEST_PATH="/etc/apollo-router/pq-manifest.json"
+REQUIRE_MANIFEST="${APOLLO_REQUIRE_PQ_MANIFEST:-false}"
+
+if [ "${REQUIRE_MANIFEST}" = "true" ]; then
+  if [ ! -f "${MANIFEST_PATH}" ]; then
+    echo "[apollo-router:dev-apps] ERROR: APOLLO_REQUIRE_PQ_MANIFEST=true but manifest not found at ${MANIFEST_PATH}"
+    exit 1
+  fi
+  echo "[apollo-router:dev-apps] PQ manifest loaded (require_manifest=true): ${MANIFEST_PATH}"
+else
+  if [ ! -f "${MANIFEST_PATH}" ]; then
+    echo "[apollo-router:dev-apps] WARNING: PQ manifest not found at ${MANIFEST_PATH} (require_manifest=false, continuing)"
+  else
+    echo "[apollo-router:dev-apps] PQ manifest loaded (require_manifest=false): ${MANIFEST_PATH}"
+  fi
+fi
+
+# 启动 router
+echo "[apollo-router:dev-apps] Starting Apollo Router on port 3000..."
+exec /dist/router --config /dist/configuration.yaml --supergraph /tmp/supergraph.graphql --hot-reload

infra/apollo-router/dev-apps-supergraph.yaml

@@ -0,0 +1,36 @@
+# Apollo Federation 2 Supergraph 组合配置（dev-apps 模式：仅启动 5 个 TS 子图）
+#
+# 适用场景：本地用 docker compose 启动 iam/classes/core-edu/content/msg/config-service/apollo-router/api-gateway/push-gateway
+# 不包含：ai（Python）、data-ana（Python）— 用户未要求启动
+#
+# 与 supergraph.yaml 的差异：
+#   - 删除 ai / data-ana 子图（避免 rover compose 因它们不可达而失败）
+#   - 所有 routing_url 用容器名（iam:3002 等）而非 host.docker.internal
+
+federation_version: =2.9.0
+
+subgraphs:
+  iam:
+    routing_url: http://iam:3002/graphql
+    schema:
+      subgraph_url: http://iam:3002/graphql
+
+  core-edu:
+    routing_url: http://core-edu:3004/graphql
+    schema:
+      subgraph_url: http://core-edu:3004/graphql
+
+  content:
+    routing_url: http://content:3005/graphql
+    schema:
+      subgraph_url: http://content:3005/graphql
+
+  msg:
+    routing_url: http://msg:3007/graphql
+    schema:
+      subgraph_url: http://msg:3007/graphql
+
+  config:
+    routing_url: http://config-service:3011/graphql
+    schema:
+      subgraph_url: http://config-service:3011/graphql

infra/apollo-router/router.yaml

@@ -17,28 +17,23 @@ supergraph:
   listen: 0.0.0.0:3000
   path: /graphql
   # 生产关闭 introspection（APOLLO_ROUTER_INTROSPECTION=false）
-  # 开发默认 true 便于调试
+  # 开发默认 true 便于调试（由 docker-compose environment 注入）
-  introspection: ${env.APOLLO_ROUTER_INTROSPECTION::true}
+  introspection: ${env.APOLLO_ROUTER_INTROSPECTION}
 
 # 持久化查询（v2.1 M3 安全加固）
 # 关联：portal-shell spec §4.3
-# - enabled: 启用 APQ（接受 hash-only 请求）
+# 注意：Apollo Router v1.45 的 persisted_queries 是 ELv2 受限特性，
-# - require_manifest: 生产设为 true，仅接受 manifest 内的 hash
+# 需连接 GraphOS（APOLLO_KEY + APOLLO_GRAPH_REF）才能启用。
-# - manifest_path: manifest 文件路径（由 portal-shell 构建产物挂载）
+# 自托管开发态禁用此特性；生产环境如需启用，需配置 GraphOS license。
-persisted_queries:
+# persisted_queries:
-  enabled: true
+#   enabled: false
-  require_manifest: ${env.APOLLO_REQUIRE_PQ_MANIFEST::false}
-  manifest_path: /etc/apollo-router/pq-manifest.json
 
 # 查询限制（v2.1 M3 安全加固）
-# 关联：portal-shell spec §5.1
+# 注意：Apollo Router v1.45 的 limits.max_depth 是 ELv2 受限特性，
-# - max_depth: 最大嵌套深度 10（防止递归攻击）
+# 需连接 GraphOS 才能启用。自托管开发态禁用；
-# - max_cost: 最大查询复杂度 1000（防止放大攻击）
+# 生产环境如需深度限制，可使用 rhai script 或 custom plugin 替代。
-# - max_batch_size: 批量查询上限 5（防止 batch 攻击）
+# limits:
-limits:
+#   max_depth: 10
-  max_depth: 10
-  max_cost: 1000
-  max_batch_size: 5
 
 # Sandbox 模式：支持开发和调试（v1.45 中 sandbox 已合并到 homepage）
 homepage:

infra/docker-compose.dev-apps.yml

@@ -0,0 +1,326 @@
+# ============================================================
+# Docker Compose Override：本地开发态应用服务（v2.1）
+#
+# 用途：将 8 个应用服务（iam/classes/core-edu/content/msg/api-gateway/push-gateway/iam）也以容器方式运行，
+# 与 infra/docker-compose.yml 中的基础设施（MySQL/Redis/Kafka/Apollo Router/portal-shell）共用 default 网络。
+#
+# 启动方式：
+#   docker compose -f infra/docker-compose.yml -f infra/docker-compose.dev-apps.yml --profile p3 up -d
+#
+# 与 infra/docker-compose.yml 的关系：
+#   - 同 name: edu-full（共用 project 命名空间与 default 网络）
+#   - 应用服务通过容器名（edu-mysql/edu-redis/edu-kafka）访问基础设施
+#   - apollo-router 已在 infra/docker-compose.yml 定义，此处不重复
+#   - config-service 已在 infra/docker-compose.yml 定义，此处不重复
+#
+# 注意：本文件只补充 docker-compose.yml 中缺失的 7 个应用服务（不含已定义的 config-service/apollo-router/portal-shell）。
+# ============================================================
+
+name: edu-full
+
+services:
+  # ============================================================
+  # IAM 服务（NestJS，端口 3002 + gRPC 50052）
+  # 依赖：MySQL（用户/角色/权限表）、Redis（缓存）、Kafka（Outbox 事件）
+  # RS256 密钥：由 infra/keys/iam-{private,public}.pem 挂载（首次启动需生成，见 scripts/gen-iam-keys.ps1）
+  # ============================================================
+  iam:
+    build:
+      context: ..
+      dockerfile: services/iam/Dockerfile
+    container_name: edu-iam
+    restart: unless-stopped
+    environment:
+      PORT: "3002"
+      GRPC_PORT: "50052"
+      DATABASE_URL: mysql://edu:${MYSQL_PASSWORD:-changeme}@edu-mysql:3306/next_edu_cloud
+      REDIS_URL: redis://edu-redis:6379
+      KAFKA_BROKERS: kafka:29092
+      KAFKA_CLIENT_ID: iam-service
+      IAM_PRIVATE_KEY_PATH: /keys/iam-private.pem
+      IAM_PUBLIC_KEY_PATH: /keys/iam-public.pem
+      JWT_ISSUER: ${JWT_ISSUER:-next-edu-cloud}
+      JWT_AUDIENCE: ${JWT_AUDIENCE:-next-edu-cloud}
+      JWT_KEY_ID: iam-rs256-v1
+      ACCESS_TOKEN_TTL: "15m"
+      REFRESH_TOKEN_TTL_DAYS: "7"
+      COOKIE_SECURE: "auto"
+      OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://edu-jaeger:4318}
+      LOG_LEVEL: ${LOG_LEVEL:-info}
+      NODE_ENV: production
+      # DEV_MODE=true 让子图 middleware 跳过鉴权，允许 apollo-router rover introspect 子图
+      # 外网请求经 api-gateway（DEV_MODE=false，RS256 验签）入站，子图在 docker 内网可信
+      DEV_MODE: "true"
+    ports:
+      - "3002:3002"
+      - "50052:50052"
+    volumes:
+      - ../infra/keys:/keys:ro
+    depends_on:
+      mysql:
+        condition: service_healthy
+      redis:
+        condition: service_started
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--spider", "http://localhost:3002/healthz"]
+      interval: 15s
+      timeout: 5s
+      start_period: 30s
+      retries: 10
+
+  # ============================================================
+  # Classes 服务（NestJS，端口 3001）
+  # 依赖：MySQL、Redis、Kafka（可选）
+  # ============================================================
+  classes:
+    build:
+      context: ..
+      dockerfile: services/classes/Dockerfile
+    container_name: edu-classes
+    restart: unless-stopped
+    environment:
+      PORT: "3001"
+      DATABASE_URL: mysql://edu:${MYSQL_PASSWORD:-changeme}@edu-mysql:3306/next_edu_cloud
+      REDIS_URL: redis://edu-redis:6379
+      KAFKA_BROKERS: kafka:29092
+      JWT_SECRET: ${JWT_SECRET:-p1-dev-secret-change-in-production}
+      JWT_ISSUER: ${JWT_ISSUER:-next-edu-cloud}
+      OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://edu-jaeger:4318}
+      LOG_LEVEL: ${LOG_LEVEL:-info}
+      NODE_ENV: production
+      # DEV_MODE=true 让子图 middleware 跳过鉴权，允许 apollo-router rover introspect 子图
+      DEV_MODE: "true"
+    ports:
+      - "3001:3001"
+    depends_on:
+      mysql:
+        condition: service_healthy
+      redis:
+        condition: service_started
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--spider", "http://localhost:3001/healthz"]
+      interval: 15s
+      timeout: 5s
+      start_period: 30s
+      retries: 10
+
+  # ============================================================
+  # Core-Edu 服务（NestJS，端口 3004 + gRPC 50054）
+  # 依赖：MySQL、Redis、Kafka（Outbox 事件）
+  # ============================================================
+  core-edu:
+    build:
+      context: ..
+      dockerfile: services/core-edu/Dockerfile
+    container_name: edu-core-edu
+    restart: unless-stopped
+    environment:
+      PORT: "3004"
+      GRPC_PORT: "50054"
+      DATABASE_URL: mysql://edu:${MYSQL_PASSWORD:-changeme}@edu-mysql:3306/next_edu_cloud
+      REDIS_URL: redis://edu-redis:6379
+      KAFKA_BROKERS: kafka:29092
+      OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://edu-jaeger:4318}
+      LOG_LEVEL: ${LOG_LEVEL:-info}
+      NODE_ENV: production
+      # DEV_MODE=true 让子图 middleware 跳过鉴权，允许 apollo-router rover introspect 子图
+      DEV_MODE: "true"
+    ports:
+      - "3004:3004"
+      - "50054:50054"
+    depends_on:
+      mysql:
+        condition: service_healthy
+      redis:
+        condition: service_started
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--spider", "http://localhost:3004/healthz"]
+      interval: 15s
+      timeout: 5s
+      start_period: 30s
+      retries: 10
+
+  # ============================================================
+  # Content 服务（NestJS，端口 3005 + gRPC 50055）
+  # 依赖：MySQL、Redis、Kafka（Outbox）、Neo4j（可选降级）、ES（可选降级）
+  # ============================================================
+  content:
+    build:
+      context: ..
+      dockerfile: services/content/Dockerfile
+    container_name: edu-content
+    restart: unless-stopped
+    environment:
+      PORT: "3005"
+      GRPC_PORT: "50055"
+      DATABASE_URL: mysql://edu:${MYSQL_PASSWORD:-changeme}@edu-mysql:3306/next_edu_cloud
+      REDIS_URL: redis://edu-redis:6379
+      KAFKA_BROKERS: kafka:29092
+      NEO4J_URL: bolt://neo4j:7687
+      NEO4J_USER: neo4j
+      NEO4J_PASSWORD: ${NEO4J_PASSWORD:-changeme}
+      ES_URL: http://elasticsearch:9200
+      OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://edu-jaeger:4318}
+      LOG_LEVEL: ${LOG_LEVEL:-info}
+      NODE_ENV: production
+      # DEV_MODE=true 让子图 middleware 跳过鉴权，允许 apollo-router rover introspect 子图
+      DEV_MODE: "true"
+    ports:
+      - "3005:3005"
+      - "50055:50055"
+    depends_on:
+      mysql:
+        condition: service_healthy
+      redis:
+        condition: service_started
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--spider", "http://localhost:3005/healthz"]
+      interval: 15s
+      timeout: 5s
+      start_period: 30s
+      retries: 10
+
+  # ============================================================
+  # Msg 服务（NestJS，端口 3007 + gRPC 50057）
+  # 依赖：MySQL、Redis、Kafka、ES（可选）、push-gateway（HTTP /internal/push）
+  # ============================================================
+  msg:
+    build:
+      context: ..
+      dockerfile: services/msg/Dockerfile
+    container_name: edu-msg
+    restart: unless-stopped
+    environment:
+      PORT: "3007"
+      GRPC_PORT: "50057"
+      DATABASE_URL: mysql://edu:${MYSQL_PASSWORD:-changeme}@edu-mysql:3306/next_edu_cloud
+      REDIS_URL: redis://edu-redis:6379
+      KAFKA_BROKERS: kafka:29092
+      ES_URL: http://elasticsearch:9200
+      PUSH_GATEWAY_URL: http://push-gateway:8081
+      PUSH_INTERNAL_TOKEN: ${PUSH_INTERNAL_TOKEN:-edu-internal-token}
+      OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://edu-jaeger:4318}
+      LOG_LEVEL: ${LOG_LEVEL:-info}
+      NODE_ENV: production
+      # DEV_MODE=true 让子图 middleware 跳过鉴权，允许 apollo-router rover introspect 子图
+      DEV_MODE: "true"
+    ports:
+      - "3007:3007"
+      - "50057:50057"
+    depends_on:
+      mysql:
+        condition: service_healthy
+      redis:
+        condition: service_started
+      push-gateway:
+        condition: service_healthy
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--spider", "http://localhost:3007/healthz"]
+      interval: 15s
+      timeout: 5s
+      start_period: 30s
+      retries: 10
+
+  # ============================================================
+  # Push-Gateway 服务（Go，端口 8081）
+  # 依赖：Redis（Pub/Sub backplane）、JWKS（iam 公钥验签）
+  # ============================================================
+  push-gateway:
+    build:
+      context: ..
+      dockerfile: services/push-gateway/Dockerfile
+    container_name: edu-push-gateway
+    restart: unless-stopped
+    environment:
+      PUSH_GATEWAY_PORT: "8081"
+      JWT_SECRET: ${JWT_SECRET:-p1-dev-secret-change-in-production}
+      REDIS_URL: redis://edu-redis:6379
+      PUSH_INTERNAL_TOKEN: ${PUSH_INTERNAL_TOKEN:-edu-internal-token}
+      JWKS_URL: http://iam:3002/v1/iam/.well-known/jwks.json
+      WS_ALLOWED_ORIGINS: ${WS_ALLOWED_ORIGINS:-http://localhost:3000,http://localhost:4000,http://localhost:4001,http://localhost:4002,http://localhost:4003,http://localhost:4010}
+      MAX_CONNS_PER_USER: "5"
+      HEARTBEAT_INTERVAL_SECONDS: "30"
+      OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://edu-jaeger:4318}
+      DEV_MODE: "false"
+    ports:
+      - "8081:8081"
+    depends_on:
+      redis:
+        condition: service_started
+    healthcheck:
+      test: ["CMD", "wget", "-q", "-O", "/dev/null", "http://localhost:8081/healthz"]
+      interval: 15s
+      timeout: 5s
+      start_period: 10s
+      retries: 5
+
+  # ============================================================
+  # API-Gateway 服务（Go，端口 8080）
+  # 依赖：iam（JWKS）、apollo-router、所有后端服务（HTTP 路由代理）
+  # ============================================================
+  api-gateway:
+    build:
+      context: ..
+      dockerfile: services/api-gateway/Dockerfile
+    container_name: edu-api-gateway
+    restart: unless-stopped
+    environment:
+      API_GATEWAY_PORT: "8080"
+      ENV: production
+      JWT_SECRET: ${JWT_SECRET:-p1-dev-secret-change-in-production}
+      JWT_ISSUER: ${JWT_ISSUER:-next-edu-cloud}
+      JWT_AUDIENCE: ${JWT_AUDIENCE:-next-edu-cloud}
+      IAM_JWKS_URL: http://iam:3002/v1/iam/.well-known/jwks.json
+      DEV_MODE: "false"
+      CLASSES_SERVICE_URL: http://classes:3001
+      IAM_SERVICE_URL: http://iam:3002
+      APOLLO_ROUTER_URL: http://apollo-router:3000
+      CORE_EDU_SERVICE_URL: http://core-edu:3004
+      CONTENT_SERVICE_URL: http://content:3005
+      MSG_SERVICE_URL: http://msg:3007
+      CORS_ORIGINS: ${CORS_ORIGINS:-http://localhost:3000,http://localhost:4000,http://localhost:4001,http://localhost:4002,http://localhost:4003,http://localhost:4010}
+      OTEL_EXPORTER_OTLP_ENDPOINT: ${OTEL_EXPORTER_OTLP_ENDPOINT:-http://edu-jaeger:4318}
+      LOG_LEVEL: ${LOG_LEVEL:-info}
+    ports:
+      - "8080:8080"
+    depends_on:
+      iam:
+        condition: service_healthy
+      apollo-router:
+        condition: service_healthy
+    healthcheck:
+      test: ["CMD", "wget", "-q", "-O", "/dev/null", "http://localhost:8080/healthz"]
+      interval: 15s
+      timeout: 5s
+      start_period: 10s
+      retries: 5
+
+  # ============================================================
+  # Apollo Router override（覆盖 docker-compose.yml 中的同名服务）
+  #
+  # 覆盖原因：默认 entrypoint.sh 会等待 ai/data-ana 子图（用户未启动），
+  # 默认 supergraph.yaml 包含 ai/data-ana 子图（rover compose 会失败）。
+  #
+  # 覆盖内容：
+  #   - 挂载 dev-apps-supergraph.yaml（仅 iam/core-edu/content/msg/config 5 个子图）
+  #   - 挂载 dev-apps-entrypoint.sh（仅等待 5 个子图）
+  #   - 添加 depends_on 5 个子图（health check 通过后才启动 router）
+  # ============================================================
+  apollo-router:
+    volumes:
+      # 追加覆盖默认 supergraph.yaml 与 entrypoint.sh
+      - ./apollo-router/dev-apps-supergraph.yaml:/dist/supergraph.yaml:ro
+      - ./apollo-router/dev-apps-entrypoint.sh:/dist/entrypoint.sh:ro
+      # 保留原 pq-manifest 挂载（来自 docker-compose.yml）
+      - ../apps/portal-shell/public/pq-manifest.json:/etc/apollo-router/pq-manifest.json:ro
+    depends_on:
+      iam:
+        condition: service_healthy
+      config-service:
+        condition: service_healthy
+      core-edu:
+        condition: service_healthy
+      content:
+        condition: service_healthy
+      msg:
+        condition: service_healthy

infra/docker-compose.yml

@@ -306,7 +306,9 @@ services:
       REDIS_URL: redis://edu-redis:6379
       KAFKA_BROKERS: kafka:29092
       ROUTER_AUTH_SECRET: ${ROUTER_AUTH_SECRET:-dev-router-secret}
-      DEV_MODE: ${DEV_MODE:-false}
+      # DEV_MODE=true 让子图 middleware 跳过鉴权，允许 apollo-router rover introspect 子图
+      DEV_MODE: "true"
+      NODE_ENV: production
       PORT: "3011"
       GRPC_PORT: "50059"
     ports:
@@ -334,6 +336,8 @@ services:
       # v2.1 M3 安全加固开关（生产设为 true，开发默认 false）
       APOLLO_REQUIRE_PQ_MANIFEST: ${APOLLO_REQUIRE_PQ_MANIFEST:-false}
       APOLLO_ROUTER_INTROSPECTION: ${APOLLO_ROUTER_INTROSPECTION:-true}
+      # Apollo Router v1.45 persisted_queries.enabled 需要布尔值（true/false）
+      APOLLO_PERSISTED_QUERIES: ${APOLLO_PERSISTED_QUERIES:-true}
     ports:
       - "3000:3000"
       - "8088:8088"
@@ -365,7 +369,8 @@ services:
       APOLLO_ROUTER_URL: http://apollo-router:3000/graphql
       NEXT_PUBLIC_API_GATEWAY_URL: http://api-gateway:8080
       NEXT_PUBLIC_REALTIME_GATEWAY_URL: http://realtime-gateway:8081
-      NEXT_PUBLIC_DEV_MODE: "true"
+      # P0-5：生产 profile 必须 false（启动校验见 src/instrumentation.ts）
+      NEXT_PUBLIC_DEV_MODE: "false"
     ports:
       - "4010:4010"
     depends_on:

services/classes/Dockerfile

@@ -1,22 +1,55 @@
 # Build stage
+# 构建上下文必须是仓库根目录（需访问 packages/shared-ts 与 pnpm-workspace.yaml）
 FROM node:22-alpine AS builder
 WORKDIR /app
 RUN corepack enable && corepack prepare pnpm@11.13.0 --activate
-COPY package.json pnpm-lock.yaml* ./
+
-# pnpm 11 默认拒绝 build scripts，独立安装时无法读取 workspace 的 allowBuilds 配置
+# 复制 workspace 配置 + tsconfig 基线
-# 使用 --ignore-scripts 跳过（@nestjs/core/protobufjs 的 postinstall 非功能必需）
+COPY package.json pnpm-lock.yaml* pnpm-workspace.yaml tsconfig.base.json ./
-# --no-frozen-lockfile：services/classes/ 无独立 pnpm-lock.yaml
+
-RUN pnpm install --no-frozen-lockfile --ignore-scripts
+# 复制 shared-proto（gRPC proto-loader 运行时加载）
-COPY tsconfig.json nest-cli.json ./
+COPY packages/shared-proto ./packages/shared-proto
-COPY src ./src
+
+# 复制 shared-ts（workspace 依赖）
+COPY packages/shared-ts ./packages/shared-ts
+
+# 复制 classes 源码
+COPY services/classes ./services/classes
+
+# HUSKY=0 跳过根 package.json 的 prepare:husky 脚本
+ENV HUSKY=0
+
+# 只安装 classes 及其 workspace 依赖（@edu/shared-ts、@edu/shared-proto）
+RUN pnpm install --no-frozen-lockfile --filter @edu/classes-service...
+
+# 构建 shared-ts（classes 依赖 @edu/shared-ts dist 产物）
+RUN cd packages/shared-ts && pnpm build
+
+# 构建 classes
+WORKDIR /app/services/classes
 RUN pnpm build
 
 # Runtime stage
 FROM node:22-alpine
 WORKDIR /app
-RUN corepack enable && corepack prepare pnpm@11.13.0 --activate
+
-COPY package.json pnpm-lock.yaml* ./
+# 复制 workspace 配置 + tsconfig 基线
-RUN pnpm install --prod --no-frozen-lockfile --ignore-scripts
+COPY package.json pnpm-lock.yaml* pnpm-workspace.yaml tsconfig.base.json ./
-COPY --from=builder /app/dist ./dist
+
+# 复制 shared-proto + shared-ts（运行时需要：proto 文件 + shared-ts dist）
+COPY packages/shared-proto ./packages/shared-proto
+COPY packages/shared-ts ./packages/shared-ts
+
+# 复制 classes 源码
+COPY services/classes ./services/classes
+
+# 从 builder 复制已编译的 prod node_modules + classes dist + shared-ts dist
+COPY --from=builder /app/services/classes/node_modules ./services/classes/node_modules
+COPY --from=builder /app/packages/shared-ts/node_modules ./packages/shared-ts/node_modules
+COPY --from=builder /app/node_modules ./node_modules
+COPY --from=builder /app/services/classes/dist ./services/classes/dist
+COPY --from=builder /app/packages/shared-ts/dist ./packages/shared-ts/dist
+
+WORKDIR /app/services/classes
 EXPOSE 3001
 CMD ["node", "dist/main.js"]

services/config-service/Dockerfile

@@ -19,7 +19,8 @@ COPY services/config-service ./services/config-service
 ENV HUSKY=0
 
 # 只安装 config-service 及其 workspace 依赖（@edu/shared-ts、@edu/shared-proto）
-RUN pnpm install --no-frozen-lockfile --filter @edu/config-service...
+# --ignore-scripts：跳过 @apollo/protobufjs 等的 postinstall（config-service 无原生模块依赖）
+RUN pnpm install --no-frozen-lockfile --ignore-scripts --filter @edu/config-service...
 
 # 构建 shared-ts（config-service 依赖 @edu/shared-ts/outbox dist 产物）
 RUN cd packages/shared-ts && pnpm build

services/content/Dockerfile

@@ -1,46 +1,56 @@
 # Build stage
+# 构建上下文必须是仓库根目录（需访问 packages/shared-ts 与 pnpm-workspace.yaml）
 FROM node:22-alpine AS builder
 WORKDIR /app
 RUN corepack enable && corepack prepare pnpm@11.13.0 --activate
 
-# 复制 workspace 根配置
+# 复制 workspace 配置 + tsconfig 基线
-COPY pnpm-workspace.yaml package.json pnpm-lock.yaml tsconfig.base.json ./
+COPY package.json pnpm-lock.yaml* pnpm-workspace.yaml tsconfig.base.json ./
 
-# 复制 shared-proto（proto 文件运行时需要）
+# 复制 shared-proto（gRPC proto-loader 运行时加载）
-COPY packages/shared-proto/package.json packages/shared-proto/
+COPY packages/shared-proto ./packages/shared-proto
-COPY packages/shared-proto/proto packages/shared-proto/proto
 
-# 复制 content 服务
+# 复制 shared-ts（workspace 依赖）
-COPY services/content/package.json services/content/tsconfig.json services/content/nest-cli.json ./services/content/
+COPY packages/shared-ts ./packages/shared-ts
-COPY services/content/src ./services/content/src
 
-# 安装依赖（不执行 scripts）
+# 复制 content 源码
-RUN pnpm install --frozen-lockfile --ignore-scripts
+COPY services/content ./services/content
 
-# 构建
+# HUSKY=0 跳过根 package.json 的 prepare:husky 脚本
-RUN cd services/content && pnpm build
+ENV HUSKY=0
+
+# 只安装 content 及其 workspace 依赖（@edu/shared-ts、@edu/shared-proto）
+# --ignore-scripts：跳过 @apollo/protobufjs 等的 postinstall（content 无原生模块依赖）
+RUN pnpm install --no-frozen-lockfile --ignore-scripts --filter @edu/content-service...
+
+# 构建 shared-ts（content 依赖 @edu/shared-ts dist 产物）
+RUN cd packages/shared-ts && pnpm build
+
+# 构建 content
+WORKDIR /app/services/content
+RUN pnpm build
 
 # Runtime stage
 FROM node:22-alpine
 WORKDIR /app
-RUN corepack enable && corepack prepare pnpm@11.13.0 --activate
 
-# 复制 workspace 配置
+# 复制 workspace 配置 + tsconfig 基线
-COPY pnpm-workspace.yaml package.json pnpm-lock.yaml tsconfig.base.json ./
+COPY package.json pnpm-lock.yaml* pnpm-workspace.yaml tsconfig.base.json ./
 
-# 复制 shared-proto（运行时加载 proto 文件）
+# 复制 shared-proto + shared-ts（运行时需要：proto 文件 + shared-ts dist）
-COPY packages/shared-proto/package.json packages/shared-proto/
+COPY packages/shared-proto ./packages/shared-proto
-COPY packages/shared-proto/proto packages/shared-proto/proto
+COPY packages/shared-ts ./packages/shared-ts
 
-# 复制 content 服务 package.json
+# 复制 content 源码
-COPY services/content/package.json ./services/content/
+COPY services/content ./services/content
 
-# 安装生产依赖
+# 从 builder 复制已编译的 prod node_modules + content dist + shared-ts dist
-RUN pnpm install --prod --frozen-lockfile --ignore-scripts
+COPY --from=builder /app/services/content/node_modules ./services/content/node_modules
-
+COPY --from=builder /app/packages/shared-ts/node_modules ./packages/shared-ts/node_modules
-# 复制构建产物
+COPY --from=builder /app/node_modules ./node_modules
 COPY --from=builder /app/services/content/dist ./services/content/dist
+COPY --from=builder /app/packages/shared-ts/dist ./packages/shared-ts/dist
 
-EXPOSE 3005 50054
 WORKDIR /app/services/content
+EXPOSE 3005 50054
 CMD ["node", "dist/main.js"]

services/core-edu/Dockerfile

@@ -1,27 +1,60 @@
-# 单阶段构建（测试用）- 使用 repo 根作为 build context
+# Build stage
-# 构建命令: docker build -t edu/core-edu:test -f services/core-edu/Dockerfile .
+# 构建上下文必须是仓库根目录（需访问 packages/shared-ts 与 pnpm-workspace.yaml）
-FROM node:22-alpine
+FROM node:22-alpine AS builder
 WORKDIR /app
-
-# 安装 pnpm
 RUN corepack enable && corepack prepare pnpm@11.13.0 --activate
 
-# 复制 workspace 配置和共享包
+# 复制 workspace 配置 + tsconfig 基线
 COPY package.json pnpm-lock.yaml* pnpm-workspace.yaml tsconfig.base.json ./
+
+# 复制 shared-proto（gRPC proto-loader 运行时加载）
 COPY packages/shared-proto ./packages/shared-proto
 
-# 复制 core-edu 服务文件
+# 复制 shared-ts（workspace 依赖）
-COPY services/core-edu/package.json services/core-edu/tsconfig.json services/core-edu/nest-cli.json ./services/core-edu/
+COPY packages/shared-ts ./packages/shared-ts
-COPY services/core-edu/src ./services/core-edu/src
 
-# 安装依赖并构建
+# 复制 core-edu 源码
+COPY services/core-edu ./services/core-edu
+
+# HUSKY=0 跳过根 package.json 的 prepare:husky 脚本
+ENV HUSKY=0
+
+# 只安装 core-edu 及其 workspace 依赖（@edu/shared-ts、@edu/shared-proto）
+# --ignore-scripts：跳过 @apollo/protobufjs 等的 postinstall（core-edu 无原生模块依赖）
+RUN pnpm install --no-frozen-lockfile --ignore-scripts --filter @edu/core-edu-service...
+
+# 构建 shared-ts（core-edu 依赖 @edu/shared-ts dist 产物）
+RUN cd packages/shared-ts && pnpm build
+
+# 构建 core-edu
 WORKDIR /app/services/core-edu
-RUN pnpm install --no-frozen-lockfile
+RUN pnpm build
-RUN npx tsc
 
 # 复制 proto 文件到 /app/proto（grpc.server.ts 的 fallback 路径）
 RUN cp -r /app/packages/shared-proto/proto /app/proto
 
-EXPOSE 3004 50053
+# Runtime stage
+FROM node:22-alpine
+WORKDIR /app
+
+# 复制 workspace 配置 + tsconfig 基线
+COPY package.json pnpm-lock.yaml* pnpm-workspace.yaml tsconfig.base.json ./
+
+# 复制 shared-proto + shared-ts（运行时需要：proto 文件 + shared-ts dist）
+COPY packages/shared-proto ./packages/shared-proto
+COPY packages/shared-ts ./packages/shared-ts
+
+# 复制 core-edu 源码
+COPY services/core-edu ./services/core-edu
+
+# 从 builder 复制已编译的 prod node_modules + core-edu dist + shared-ts dist + proto
+COPY --from=builder /app/services/core-edu/node_modules ./services/core-edu/node_modules
+COPY --from=builder /app/packages/shared-ts/node_modules ./packages/shared-ts/node_modules
+COPY --from=builder /app/node_modules ./node_modules
+COPY --from=builder /app/services/core-edu/dist ./services/core-edu/dist
+COPY --from=builder /app/packages/shared-ts/dist ./packages/shared-ts/dist
+COPY --from=builder /app/proto ./proto
+
 WORKDIR /app/services/core-edu
+EXPOSE 3004 50053
 CMD ["node", "dist/main.js"]