life-echo/api/development.sh

#!/usr/bin/env bash

set -euo pipefail

RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m'

ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
REPO_ROOT="$(cd "${ROOT_DIR}/.." && pwd)"
EVAL_WEB_DIR="${REPO_ROOT}/app-eval-web"
VENV_DIR="${ROOT_DIR}/.venv"
PYTHON_BIN="${VENV_DIR}/bin/python"
UVICORN_BIN="${VENV_DIR}/bin/uvicorn"
CELERY_BIN="${VENV_DIR}/bin/celery"

# 本地全栈：默认可在 API 启动时 purge Celery 队列；生产请勿使用此脚本
export APP_ENV="${APP_ENV:-development}"
export CELERY_PURGE_BROKER_ON_STARTUP="${CELERY_PURGE_BROKER_ON_STARTUP:-1}"

API_HOST="${API_HOST:-0.0.0.0}"
API_PORT="${API_PORT:-8000}"
CELERY_POOL="${CELERY_POOL:-solo}"
SKIP_INSTALL="${SKIP_INSTALL:-0}"
SKIP_INFRA="${SKIP_INFRA:-0}"
SHUTDOWN_TIMEOUT="${SHUTDOWN_TIMEOUT:-12}"

# 由 internal-eval.sh 开启：在 main:app + Celery 之外再启 internal_main(:8001) 与 app-eval-web
LIFE_ECHO_WITH_INTERNAL_EVAL="${LIFE_ECHO_WITH_INTERNAL_EVAL:-0}"
# 若 :8000 已由其他 development 实例占用，仅附加 :8001 + 前端（需自备同一份 Celery/主站）
EVAL_ATTACH_ONLY="${EVAL_ATTACH_ONLY:-0}"
INTERNAL_EVAL_HOST="${INTERNAL_EVAL_HOST:-0.0.0.0}"
INTERNAL_EVAL_PORT="${INTERNAL_EVAL_PORT:-7999}"
START_EVAL_WEB="${START_EVAL_WEB:-1}"
OPEN_EVAL_WEB="${OPEN_EVAL_WEB:-1}"
EVAL_WEB_PORT="${EVAL_WEB_PORT:-5174}"

API_PID=""
CELERY_PID=""
INTERNAL_EVAL_PID=""
EVAL_WEB_PID=""
CLEANED_UP=0
INFRA_STARTED=0

print_header() {
  echo -e "\n${BLUE}========================================${NC}"
  echo -e "${BLUE}$1${NC}"
  echo -e "${BLUE}========================================${NC}"
}

print_ok() {
  echo -e "${GREEN}✓ $1${NC}"
}

print_warn() {
  echo -e "${YELLOW}⚠ $1${NC}"
}

print_err() {
  echo -e "${RED}✗ $1${NC}"
}

is_pid_alive() {
  local pid="$1"
  [[ -n "${pid}" ]] && kill -0 "${pid}" 2>/dev/null
}

wait_pid_exit() {
  local pid="$1"
  local timeout="$2"
  local waited=0

  while is_pid_alive "${pid}"; do
    if (( waited >= timeout )); then
      return 1
    fi
    sleep 1
    waited=$((waited + 1))
  done
  return 0
}

kill_children_term() {
  local pid="$1"
  local children

  children="$(pgrep -P "${pid}" 2>/dev/null || true)"
  if [[ -n "${children}" ]]; then
    # 先递归处理子进程，避免 reloader/server 残留
    while IFS= read -r child_pid; do
      [[ -z "${child_pid}" ]] && continue
      kill_children_term "${child_pid}"
      kill -TERM "${child_pid}" 2>/dev/null || true
    done <<< "${children}"
  fi
}

stop_process_gracefully() {
  local name="$1"
  local pid="$2"
  local timeout="${3:-10}"

  if ! is_pid_alive "${pid}"; then
    print_ok "${name} 已退出"
    return 0
  fi

  print_warn "正在停止 ${name}（PID: ${pid}）..."
  kill_children_term "${pid}"
  kill -TERM "${pid}" 2>/dev/null || true

  if wait_pid_exit "${pid}" "${timeout}"; then
    print_ok "${name} 已停止"
    return 0
  fi

  print_warn "${name} 在 ${timeout}s 内未退出，准备强制结束"
  kill -KILL "${pid}" 2>/dev/null || true
  wait_pid_exit "${pid}" 3 || true
  print_ok "${name} 已强制结束"
}

cleanup() {
  if [[ "${CLEANED_UP}" == "1" ]]; then
    return 0
  fi
  CLEANED_UP=1

  print_header "正在关闭开发环境"

  if is_pid_alive "${EVAL_WEB_PID}"; then
    stop_process_gracefully "eval-web (Vite)" "${EVAL_WEB_PID}" "${SHUTDOWN_TIMEOUT}"
  fi

  if is_pid_alive "${INTERNAL_EVAL_PID}"; then
    stop_process_gracefully "Internal Eval API (:${INTERNAL_EVAL_PORT})" "${INTERNAL_EVAL_PID}" "${SHUTDOWN_TIMEOUT}"
  fi

  if is_pid_alive "${API_PID}"; then
    stop_process_gracefully "FastAPI" "${API_PID}" "${SHUTDOWN_TIMEOUT}"
  fi

  if is_pid_alive "${CELERY_PID}"; then
    stop_process_gracefully "Celery" "${CELERY_PID}" "${SHUTDOWN_TIMEOUT}"
  fi

  if [[ "${INFRA_STARTED}" == "1" ]]; then
    print_warn "正在停止 PostgreSQL / Redis 容器..."
    (
      cd "${ROOT_DIR}" && docker compose -f docker-compose.dev.yml stop
    ) >/dev/null 2>&1 || true
    print_ok "PostgreSQL/Redis 容器已停止"
  fi
}

require_cmd() {
  local cmd="$1"
  if ! command -v "${cmd}" >/dev/null 2>&1; then
    print_err "未找到命令: ${cmd}"
    exit 1
  fi
}

start_infra() {
  print_header "启动 PostgreSQL 和 Redis"
  cd "${ROOT_DIR}"
  docker compose -f docker-compose.dev.yml up -d
  INFRA_STARTED=1
  print_ok "PostgreSQL 127.0.0.1:48291，Redis 127.0.0.1:48307（见 docker-compose.dev.yml / .env.example）"
  print_ok "基础设施已就绪"
}

# Docker 刚启动时 Postgres 可能尚未接受连接，立即跑 Alembic 会误报失败
wait_postgres_ready() {
  local retries=30
  local i=0
  print_header "等待 PostgreSQL 就绪"
  cd "${ROOT_DIR}"
  while (( i < retries )); do
    if docker compose -f docker-compose.dev.yml exec -T postgres \
      pg_isready -U postgres >/dev/null 2>&1; then
      print_ok "PostgreSQL 已就绪"
      return 0
    fi
    sleep 1
    i=$((i + 1))
  done
  print_warn "PostgreSQL 在 ${retries}s 内未就绪，迁移可能失败"
  return 1
}

get_effective_database_url() {
  if [[ -n "${DATABASE_URL:-}" ]]; then
    printf '%s\n' "${DATABASE_URL}"
    return 0
  fi

  if [[ -f "${ROOT_DIR}/.env" ]]; then
    local line
    line="$(sed -n 's/^DATABASE_URL=//p' "${ROOT_DIR}/.env" | sed -n '1p')"
    line="${line%\"}"
    line="${line#\"}"
    line="${line%\'}"
    line="${line#\'}"
    if [[ -n "${line}" ]]; then
      printf '%s\n' "${line}"
      return 0
    fi
  fi

  return 1
}

get_effective_redis_url() {
  if [[ -n "${REDIS_URL:-}" ]]; then
    printf '%s\n' "${REDIS_URL}"
    return 0
  fi

  if [[ -f "${ROOT_DIR}/.env" ]]; then
    local line
    line="$(sed -n 's/^REDIS_URL=//p' "${ROOT_DIR}/.env" | sed -n '1p')"
    line="${line%\"}"
    line="${line#\"}"
    line="${line%\'}"
    line="${line#\'}"
    if [[ -n "${line}" ]]; then
      printf '%s\n' "${line}"
      return 0
    fi
  fi

  return 1
}

extract_url_port() {
  local url="$1"
  local default_port="$2"

  if [[ "${url}" =~ :([0-9]+)(/|\?|$) ]]; then
    printf '%s\n' "${BASH_REMATCH[1]}"
    return 0
  fi

  printf '%s\n' "${default_port}"
}

wait_host_infra_ready() {
  local database_url redis_url pg_port redis_port

  if ! database_url="$(get_effective_database_url)"; then
    print_warn "无法解析 DATABASE_URL，跳过宿主机 PostgreSQL 端口检查"
  else
    pg_port="$(extract_url_port "${database_url}" "5432")"
    if wait_for_tcp_listener "$$" "${pg_port}" 12; then
      print_ok "宿主机 PostgreSQL 端口已监听 (:${pg_port})"
    else
      print_warn "宿主机 PostgreSQL 端口未监听 (:${pg_port})；请检查 .env 与 docker-compose.dev.yml 端口映射"
    fi
  fi

  if ! redis_url="$(get_effective_redis_url)"; then
    print_warn "无法解析 REDIS_URL，跳过宿主机 Redis 端口检查"
  else
    redis_port="$(extract_url_port "${redis_url}" "6379")"
    if wait_for_tcp_listener "$$" "${redis_port}" 12; then
      print_ok "宿主机 Redis 端口已监听 (:${redis_port})"
    else
      print_warn "宿主机 Redis 端口未监听 (:${redis_port})；请检查 .env 与 docker-compose.dev.yml 端口映射"
    fi
  fi
}

warn_database_url_host_pitfall() {
  local database_url
  local host

  if ! database_url="$(get_effective_database_url)"; then
    return 0
  fi

  if [[ "${database_url}" =~ @([^:/?#]+) ]]; then
    host="${BASH_REMATCH[1]}"
    case "${host}" in
      postgres|db|postgres-dev|postgresql)
        print_warn "检测到 DATABASE_URL 主机为 ${host}；在宿主机执行 Alembic/uvicorn 时通常应使用 localhost"
        ;;
    esac
  fi
}

print_alembic_failure_hint() {
  local log_file="$1"
  local log_output

  log_output="$(sed -n '1,200p' "${log_file}")"
  if [[ "${log_output}" == *'could not translate host name "postgres"'* ]] || [[ "${log_output}" == *"Name or service not known"* ]]; then
    print_warn "看起来 DATABASE_URL 指向了容器内主机名；在宿主机运行时请改用 localhost:48291（见 docker-compose.dev.yml）"
  elif [[ "${log_output}" == *"Connection refused"* ]] || [[ "${log_output}" == *"could not connect to server"* ]]; then
    print_warn "PostgreSQL 连接被拒绝；请确认容器已启动且 DATABASE_URL 与 docker-compose.dev.yml 暴露端口一致"
  elif [[ "${log_output}" == *"password authentication failed"* ]]; then
    print_warn "PostgreSQL 用户名或密码不匹配；请核对 .env.development 中的 DATABASE_URL"
  elif [[ "${log_output}" == *"No such file or directory"* ]] || [[ "${log_output}" == *"can't open file"* ]]; then
    print_warn "Alembic 依赖的文件或工作目录可能不正确；请确认在 api/ 目录运行脚本"
  fi
}

is_port_listening() {
  local port="$1"

  if command -v lsof >/dev/null 2>&1; then
    lsof -nP -iTCP:"${port}" -sTCP:LISTEN >/dev/null 2>&1
    return $?
  fi

  if [[ -x "${PYTHON_BIN}" ]]; then
    "${PYTHON_BIN}" - "${port}" <<'PY' >/dev/null 2>&1
import socket
import sys

sock = socket.socket()
sock.settimeout(0.2)
try:
    sock.connect(("127.0.0.1", int(sys.argv[1])))
except OSError:
    raise SystemExit(1)
finally:
    sock.close()
raise SystemExit(0)
PY
    return $?
  fi

  return 1
}

wait_for_tcp_listener() {
  local pid="$1"
  local port="$2"
  local timeout="${3:-8}"
  local waited=0

  while (( waited < timeout )); do
    if is_port_listening "${port}"; then
      return 0
    fi
    if ! is_pid_alive "${pid}"; then
      return 1
    fi
    sleep 1
    waited=$((waited + 1))
  done

  return 2
}

ensure_background_process_alive() {
  local name="$1"
  local pid="$2"

  sleep 1
  if ! is_pid_alive "${pid}"; then
    print_err "${name} 启动后立即退出，请查看上方日志"
    exit 1
  fi
}

ensure_venv() {
  print_header "检查 Python 虚拟环境"

  if [[ ! -d "${VENV_DIR}" ]]; then
    print_warn ".venv 不存在，正在创建"
    uv venv "${VENV_DIR}"
  fi

  if [[ "${SKIP_INSTALL}" != "1" ]]; then
    print_header "安装 Python 依赖"
    uv sync
    print_ok "依赖安装完成"
  else
    print_warn "已跳过依赖安装 (SKIP_INSTALL=1)"
  fi
}

# 本地约定：以 .env.development 为真源；每次一键启动都从 .env.development 覆盖 .env，供 pydantic Settings(env_file=".env") 读取。
# 请勿仅在 .env 里改密钥而不同步回 .env.development，否则下次启动会被覆盖。
ensure_dotenv_from_development() {
  print_header "准备本地 .env"
  if [[ -f "${ROOT_DIR}/.env.development" ]]; then
    cp "${ROOT_DIR}/.env.development" "${ROOT_DIR}/.env"
    print_ok "已从 .env.development 同步为 .env"
    return 0
  fi
  print_warn "未找到 .env.development，无法自动生成 .env"
  print_warn "请执行: cp api/.env.example api/.env.development 后按说明填写，再运行 ./development.sh"
}

check_env_file() {
  print_header "检查环境变量文件"
  if [[ ! -f "${ROOT_DIR}/.env" ]]; then
    print_warn "未找到 .env，应用可能因缺少配置启动失败"
  else
    print_ok "检测到 .env"
    warn_database_url_host_pitfall
  fi
}

run_migrations() {
  print_header "执行数据库迁移"
  cd "${ROOT_DIR}"
  local log_file
  log_file="$(mktemp -t life-echo-alembic.XXXXXX.log)"

  if uv run alembic upgrade head >"${log_file}" 2>&1; then
    print_ok "Alembic 迁移已就绪"
    rm -f "${log_file}"
  else
    print_warn "Alembic 迁移失败（可能数据库未启动或 DATABASE_URL 未配置），应用启动可能失败"
    print_alembic_failure_hint "${log_file}"
    print_warn "Alembic 输出（最近 40 行）："
    tail -n 40 "${log_file}"
    rm -f "${log_file}"
  fi
}

check_internal_eval_key() {
  print_header "检查内部评测密钥"
  if [[ -f "${ROOT_DIR}/.env" ]] && grep -qE '^INTERNAL_EVAL_API_KEY=.+' "${ROOT_DIR}/.env" 2>/dev/null; then
    print_ok "已在 .env 中配置 INTERNAL_EVAL_API_KEY"
    return 0
  fi
  if [[ -n "${INTERNAL_EVAL_API_KEY:-}" ]]; then
    print_ok "已从环境变量传入 INTERNAL_EVAL_API_KEY"
    return 0
  fi
  print_err "未配置 INTERNAL_EVAL_API_KEY：内部评测接口将返回 503。"
  print_err "请在 api/.env.development（或 .env）中加入一行，例如："
  print_err "  INTERNAL_EVAL_API_KEY=\"your-long-random-secret\""
  exit 1
}

start_eval_web() {
  print_header "启动 app-eval-web (Vite)"
  if [[ ! -d "${EVAL_WEB_DIR}" ]]; then
    print_err "未找到 ${EVAL_WEB_DIR}"
    exit 1
  fi
  if [[ ! -d "${EVAL_WEB_DIR}/node_modules" ]]; then
    print_err "请先执行: cd app-eval-web && npm install"
    exit 1
  fi
  require_cmd "npm"

  local api_key="${INTERNAL_EVAL_API_KEY:-}"
  if [[ -z "${api_key}" ]] && [[ -f "${ROOT_DIR}/.env" ]]; then
    api_key="$(grep -E '^INTERNAL_EVAL_API_KEY=' "${ROOT_DIR}/.env" | head -1 | cut -d= -f2- | tr -d '\r' | sed 's/^"//;s/"$//')"
  fi
  if [[ -z "${api_key}" ]]; then
    print_err "无法解析 INTERNAL_EVAL_API_KEY，无法为 Vite 注入 VITE_EVAL_API_KEY"
    exit 1
  fi

  local vite_extra=()
  if [[ "${OPEN_EVAL_WEB}" == "1" ]]; then
    vite_extra+=(--open)
  fi

  (
    cd "${EVAL_WEB_DIR}"
    VITE_EVAL_API_KEY="${api_key}" \
    VITE_EVAL_PROXY_TARGET="http://127.0.0.1:${INTERNAL_EVAL_PORT}" \
      npm run dev -- --host 127.0.0.1 --port "${EVAL_WEB_PORT}" "${vite_extra[@]}"
  ) &
  EVAL_WEB_PID=$!
  print_ok "eval-web 已启动 (PID: ${EVAL_WEB_PID}) → http://127.0.0.1:${EVAL_WEB_PORT}/"
}

start_internal_eval_http() {
  check_internal_eval_key
  print_header "启动内部评测 API (internal_main :${INTERNAL_EVAL_PORT})"
  cd "${ROOT_DIR}"

  if is_port_listening "${INTERNAL_EVAL_PORT}"; then
    print_err "端口 ${INTERNAL_EVAL_PORT} 已被占用，无法启动内部评测 Uvicorn。"
    print_err "请先结束占用进程，或设置 INTERNAL_EVAL_PORT 为其他端口"
    exit 1
  fi

  "${UVICORN_BIN}" app.internal_main:internal_app --reload \
    --reload-exclude 'alembic/**' \
    --reload-exclude 'alembic.ini' \
    --host "${INTERNAL_EVAL_HOST}" --port "${INTERNAL_EVAL_PORT}" &
  INTERNAL_EVAL_PID=$!
  local api_start_status=0
  if wait_for_tcp_listener "${INTERNAL_EVAL_PID}" "${INTERNAL_EVAL_PORT}" 8; then
    api_start_status=0
  else
    api_start_status=$?
  fi

  case "${api_start_status}" in
    0)
      print_ok "Internal Eval API 已启动 (PID: ${INTERNAL_EVAL_PID})"
      ;;
    1)
      print_err "Internal Eval API 启动失败，进程已退出；请查看上方 Uvicorn 日志"
      exit 1
      ;;
    *)
      print_err "Internal Eval API 进程仍存活，但端口 ${INTERNAL_EVAL_PORT} 未在预期时间内开始监听"
      exit 1
      ;;
  esac

  if [[ "${START_EVAL_WEB}" == "1" ]]; then
    start_eval_web
  else
    print_warn "已跳过 eval-web (START_EVAL_WEB=0)"
  fi

  echo
  echo -e "${BLUE}── 内部评测${NC}"
  echo "评测 Web UI:  http://127.0.0.1:${EVAL_WEB_PORT}/ （Vite /internal → :${INTERNAL_EVAL_PORT}）"
  echo "内部评测 API: http://127.0.0.1:${INTERNAL_EVAL_PORT}/health"
  echo "评测 REST:    http://127.0.0.1:${INTERNAL_EVAL_PORT}/internal/api/evaluation"
  if [[ "${INTERNAL_EVAL_ENABLE_DOCS:-}" == "1" ]] || grep -qE '^INTERNAL_EVAL_ENABLE_DOCS=true' "${ROOT_DIR}/.env" 2>/dev/null; then
    echo "内部评测文档: http://127.0.0.1:${INTERNAL_EVAL_PORT}/docs"
  fi
  echo "说明: api/docs/internal-eval.md"
}

start_services() {
  print_header "启动 FastAPI 和 Celery"
  cd "${ROOT_DIR}"

  local skip_main=0
  if [[ "${EVAL_ATTACH_ONLY}" == "1" ]] && is_port_listening "${API_PORT}"; then
    skip_main=1
    print_warn "EVAL_ATTACH_ONLY=1：:${API_PORT} 已在监听，跳过本脚本内的主站与 Celery（请确保别处已有 Worker）"
  fi

  if [[ "${skip_main}" == "1" ]] && [[ "${LIFE_ECHO_WITH_INTERNAL_EVAL}" != "1" ]]; then
    print_err "EVAL_ATTACH_ONLY=1 仅用于在已有主站时附加内部评测；请使用 ./internal-eval.sh 或导出 LIFE_ECHO_WITH_INTERNAL_EVAL=1"
    exit 1
  fi

  if [[ "${skip_main}" == "0" ]]; then
    if is_port_listening "${API_PORT}"; then
      print_err "端口 ${API_PORT} 已被占用，无法启动新的 Uvicorn。"
      print_err "请先结束占用进程，例如: lsof -nP -iTCP:${API_PORT} -sTCP:LISTEN"
      print_err "若主站已在其他终端由本脚本启动，可改用: EVAL_ATTACH_ONLY=1 ./development.sh"
      exit 1
    fi

    "${UVICORN_BIN}" main:app --reload \
      --reload-exclude 'alembic/**' \
      --reload-exclude 'alembic.ini' \
      --host "${API_HOST}" --port "${API_PORT}" &
    API_PID=$!
    local api_start_status=0
    if wait_for_tcp_listener "${API_PID}" "${API_PORT}" 8; then
      api_start_status=0
    else
      api_start_status=$?
    fi

    case "${api_start_status}" in
      0)
        print_ok "FastAPI 已启动 (PID: ${API_PID})"
        ;;
      1)
        print_err "FastAPI 启动失败，进程已退出；请查看上方 Uvicorn 日志"
        exit 1
        ;;
      *)
        print_err "FastAPI 进程仍存活，但端口 ${API_PORT} 未在预期时间内开始监听"
        exit 1
        ;;
    esac

    "${CELERY_BIN}" -A app.tasks.celery_app worker --loglevel=info --pool="${CELERY_POOL}" -Q celery,memory_idle &
    CELERY_PID=$!
    ensure_background_process_alive "Celery" "${CELERY_PID}"
    print_ok "Celery 已启动 (PID: ${CELERY_PID})"
  fi

  if [[ "${LIFE_ECHO_WITH_INTERNAL_EVAL}" == "1" ]]; then
    start_internal_eval_http
  fi

  echo
  echo -e "${GREEN}开发环境启动完成${NC}"
  if [[ -n "${API_PID}" ]]; then
    echo "主站文档: http://localhost:${API_PORT}/docs"
    echo "健康检查: http://localhost:${API_PORT}/health"
  fi
  echo "按 Ctrl+C 停止所有进程"
}

main() {
  if [[ "${LIFE_ECHO_WITH_INTERNAL_EVAL}" == "1" ]]; then
    print_header "Life Echo 开发环境 + 内部评测（主站 + :${INTERNAL_EVAL_PORT} + Eval Web）"
  else
    print_header "Life Echo 开发环境一键启动"
  fi

  require_cmd "uv"
  if [[ "${SKIP_INFRA}" != "1" ]]; then
    require_cmd "docker"
  fi

  trap cleanup EXIT INT TERM

  if [[ "${SKIP_INFRA}" != "1" ]]; then
    start_infra
    wait_postgres_ready || true
  else
    print_warn "已跳过 docker 基础设施 (SKIP_INFRA=1)"
  fi

  ensure_venv
  ensure_dotenv_from_development
  check_env_file
  wait_host_infra_ready
  run_migrations
  start_services

  local wait_pids=()
  [[ -n "${API_PID}" ]] && wait_pids+=("${API_PID}")
  [[ -n "${CELERY_PID}" ]] && wait_pids+=("${CELERY_PID}")
  [[ -n "${INTERNAL_EVAL_PID}" ]] && wait_pids+=("${INTERNAL_EVAL_PID}")
  [[ -n "${EVAL_WEB_PID}" ]] && wait_pids+=("${EVAL_WEB_PID}")
  if (( ${#wait_pids[@]} == 0 )); then
    print_err "没有可等待的进程，退出"
    exit 1
  fi
  wait "${wait_pids[@]}"
}

main "$@"