start-space.sh

#!/usr/bin/env sh
set -eu

# ============================================================
# 1. 管理员密码来源
# ============================================================
if [ -n "${ADMIN_PASSWORD_V2:-}" ]; then
    ADMIN_PASSWORD="$ADMIN_PASSWORD_V2"
    echo "Admin password secret source: ADMIN_PASSWORD_V2"
else
    echo "Admin password secret source: ADMIN_PASSWORD"
fi

# ============================================================
# 2. 自部署 PostgreSQL 初始化与启动
# ============================================================
init_postgres() {
    echo "Initializing and starting self-deployed PostgreSQL..."

    # 如果数据目录不存在，则执行初始化
    if [ ! -d "$PGDATA" ]; then
        echo "PostgreSQL data directory not found, initializing..."
        mkdir -p "$PGDATA"
        chown postgres:postgres "$PGDATA"

        # 使用临时文件传递密码
        pwfile=$(mktemp)
        echo "$POSTGRES_PASSWORD" > "$pwfile"
        chown postgres:postgres "$pwfile"
        chmod 600 "$pwfile"
        su-exec postgres initdb -D "$PGDATA" --auth=scram-sha-256 --pwfile="$pwfile"
        rm -f "$pwfile"
    fi

    # 配置 postgresql.conf
    echo "listen_addresses = 'localhost'" >> "$PGDATA/postgresql.conf"
    echo "port = 5432" >> "$PGDATA/postgresql.conf"
    echo "unix_socket_directories = '/tmp'" >> "$PGDATA/postgresql.conf"

    # 启动 PostgreSQL
    echo "Starting PostgreSQL daemon..."
    su-exec postgres pg_ctl -D "$PGDATA" -o "-c unix_socket_directories=/tmp" start

    # 等待就绪
    echo "Waiting for PostgreSQL to accept connections..."
    for i in $(seq 1 30); do
        if su-exec postgres pg_isready -U postgres -h localhost -p 5432 >/dev/null 2>&1; then
            echo "PostgreSQL is ready."
            break
        fi
        sleep 1
    done

    # 创建应用用户和数据库（如果不存在）
    echo "Ensuring application user and database exist..."
    export PGPASSWORD="$POSTGRES_PASSWORD"
    # 取消可能存在的 PGDATABASE 环境变量，并显式指定数据库为 postgres
    unset PGDATABASE
    su-exec postgres psql -h localhost -p 5432 -d postgres -v ON_ERROR_STOP=1 --username postgres <<-EOSQL
        DO \$\$
        BEGIN
            IF NOT EXISTS (SELECT FROM pg_roles WHERE rolname = '$POSTGRES_USER') THEN
                CREATE ROLE $POSTGRES_USER WITH LOGIN PASSWORD '$POSTGRES_PASSWORD';
            END IF;
        END
        \$\$;
        SELECT 'CREATE DATABASE $POSTGRES_DB OWNER $POSTGRES_USER'
        WHERE NOT EXISTS (SELECT FROM pg_database WHERE datname = '$POSTGRES_DB')\gexec
EOSQL
    unset PGPASSWORD

    # 设置 DATABASE_URL 及拆分变量
    export DATABASE_URL="postgresql://${POSTGRES_USER}:${POSTGRES_PASSWORD}@localhost:5432/${POSTGRES_DB}?sslmode=disable"
    export DATABASE_HOST=localhost
    export DATABASE_PORT=5432
    export DATABASE_USER=$POSTGRES_USER
    export DATABASE_PASSWORD=$POSTGRES_PASSWORD
    export DATABASE_DBNAME=$POSTGRES_DB
    export DATABASE_SSLMODE=disable

    echo "PostgreSQL self-deployment ready. DATABASE_URL set to local instance."
}

# ============================================================
# 3. 数据库连接解析（优先使用外部连接）
# ============================================================
if [ -n "${EXTERNAL_DATABASE_URL:-}" ]; then
    echo "EXTERNAL_DATABASE_URL provided, using external database instead of self-deployed."
    export DATABASE_URL="$EXTERNAL_DATABASE_URL"
    eval $(python3 -c "
import os, urllib.parse
url = os.environ['DATABASE_URL']
parsed = urllib.parse.urlparse(url)
user = parsed.username or ''
password = parsed.password or ''
host = parsed.hostname or ''
port = str(parsed.port or 5432)
dbname = parsed.path.lstrip('/') or ''
query = urllib.parse.parse_qs(parsed.query)
sslmode = query.get('sslmode', ['require'])[0]
print(f'export DATABASE_HOST={host}')
print(f'export DATABASE_PORT={port}')
print(f'export DATABASE_USER={user}')
print(f'export DATABASE_PASSWORD={password}')
print(f'export DATABASE_DBNAME={dbname}')
print(f'export DATABASE_SSLMODE={sslmode}')
")
    echo "Parsed external database: DATABASE_HOST=$DATABASE_HOST, DATABASE_DBNAME=$DATABASE_DBNAME"
else
    init_postgres
fi

# ============================================================
# 4. 检查必需的数据库变量
# ============================================================
required_db_vars="DATABASE_HOST DATABASE_PORT DATABASE_USER DATABASE_DBNAME"
for var in $required_db_vars; do
    eval "val=\${$var:-}"
    if [ -z "$val" ]; then
        echo "Missing required database variable: $var. Please set DATABASE_URL or EXTERNAL_DATABASE_URL." >&2
        exit 1
    fi
done

# ============================================================
# 5. 检查其他必需密钥
# ============================================================
required_vars="ADMIN_PASSWORD JWT_SECRET TOTP_ENCRYPTION_KEY"
for var in $required_vars; do
    eval "val=\${$var:-}"
    if [ -z "$val" ]; then
        echo "Missing required secret: $var" >&2
        exit 1
    fi
done

# ============================================================
# 6. 默认值设置
# ============================================================
ADMIN_EMAIL="${ADMIN_EMAIL:-2691539771@qq.com}"
SERVER_PORT="${SERVER_PORT:-7860}"
DATA_DIR="${DATA_DIR:-/tmp/gateway-data}"
DATABASE_SSLMODE="${DATABASE_SSLMODE:-require}"
REDIS_HOST="${REDIS_HOST:-127.0.0.1}"
REDIS_PORT="${REDIS_PORT:-6379}"
REDIS_PASSWORD="${REDIS_PASSWORD:-}"
REDIS_DB="${REDIS_DB:-0}"

# ============================================================
# 7. 创建目录
# ============================================================
mkdir -p "$DATA_DIR" /tmp/gateway-redis
mkdir -p "$(dirname "$PGDATA")"
chown -R sub2api:sub2api "$DATA_DIR" /tmp/gateway-redis

# ============================================================
# 8. 启动 Redis
# ============================================================
echo "Starting Redis on ${REDIS_HOST}:${REDIS_PORT}..."
redis-server --daemonize yes \
    --bind "${REDIS_HOST}" \
    --port "${REDIS_PORT}" \
    --dir /tmp/gateway-redis \
    --save "" \
    --appendonly no \
    ${REDIS_PASSWORD:+--requirepass "$REDIS_PASSWORD"}

echo "Waiting for Redis..."
while ! redis-cli -h "$REDIS_HOST" -p "$REDIS_PORT" ${REDIS_PASSWORD:+-a "$REDIS_PASSWORD"} ping >/dev/null 2>&1; do
    sleep 1
done
echo "Redis is ready."

# ============================================================
# 9. 导出环境变量供子进程使用
# ============================================================
export AUTO_SETUP="${AUTO_SETUP:-true}"
export ADMIN_EMAIL ADMIN_PASSWORD JWT_SECRET TOTP_ENCRYPTION_KEY
export DATA_DIR
export DATABASE_HOST DATABASE_PORT DATABASE_USER DATABASE_PASSWORD DATABASE_DBNAME DATABASE_SSLMODE
export REDIS_HOST REDIS_PORT REDIS_PASSWORD REDIS_DB
export SERVER_HOST="${SERVER_HOST:-0.0.0.0}" SERVER_PORT
export SERVER_MODE="${SERVER_MODE:-release}"
export PORT="${PORT:-$SERVER_PORT}"
GIN_MODE="${GIN_MODE:-release}"

# ============================================================
# 10. 辅助函数
# ============================================================
migration_count() {
    PGPASSWORD="${DATABASE_PASSWORD:-}" psql \
        -h "$DATABASE_HOST" -p "$DATABASE_PORT" \
        -U "$DATABASE_USER" -d "$DATABASE_DBNAME" \
        -tAc "select count(1) from schema_migrations;" 2>/dev/null || printf '0\n'
}

sync_balance_cache_once() {
    [ "${BALANCE_CACHE_SYNC_ENABLED:-true}" != "true" ] && return 0
    rows="$(PGPASSWORD="${DATABASE_PASSWORD:-}" psql -h "$DATABASE_HOST" -p "$DATABASE_PORT" -U "$DATABASE_USER" -d "$DATABASE_DBNAME" -qAt -F '|' -c "select id, balance::double precision from users where deleted_at is null and status = 'active';" 2>/tmp/balance-cache-sync.err || true)"
    [ -z "$rows" ] && return 0
    ttl="${BALANCE_CACHE_SYNC_TTL_SECONDS:-300}"
    printf '%s\n' "$rows" | while IFS='|' read -r user_id balance; do
        case "$user_id" in ''|*[!0-9]*) continue ;; esac
        redis-cli -h "$REDIS_HOST" -p "$REDIS_PORT" ${REDIS_PASSWORD:+-a "$REDIS_PASSWORD"} -n "$REDIS_DB" \
            SETEX "billing:balance:${user_id}" "$ttl" "$balance" >/dev/null 2>&1 || true
    done
}

start_balance_cache_sync() {
    [ "${BALANCE_CACHE_SYNC_ENABLED:-true}" != "true" ] && return 0
    ( interval="${BALANCE_CACHE_SYNC_INTERVAL_SECONDS:-60}"
    while :; do
        sync_balance_cache_once || true
        sleep "$interval"
    done ) &
    echo "Balance cache sync loop started."
}

write_runtime_config() {
    cat >"$DATA_DIR/config.yaml" <<EOF
server:
  host: "${SERVER_HOST}"
  port: ${SERVER_PORT}
  mode: "${SERVER_MODE}"
database:
  host: "${DATABASE_HOST}"
  port: ${DATABASE_PORT}
  user: "${DATABASE_USER}"
  password: "${DATABASE_PASSWORD}"
  dbname: "${DATABASE_DBNAME}"
  sslmode: "${DATABASE_SSLMODE}"
  max_open_conns: ${DATABASE_MAX_OPEN_CONNS:-10}
  max_idle_conns: ${DATABASE_MAX_IDLE_CONNS:-4}
  conn_max_lifetime_minutes: ${DATABASE_CONN_MAX_LIFETIME_MINUTES:-10}
  conn_max_idle_time_minutes: ${DATABASE_CONN_MAX_IDLE_TIME_MINUTES:-5}
redis:
  host: "${REDIS_HOST}"
  port: ${REDIS_PORT}
  password: "${REDIS_PASSWORD:-}"
  db: ${REDIS_DB:-0}
  enable_tls: ${REDIS_ENABLE_TLS:-false}
jwt:
  secret: "${JWT_SECRET}"
  expire_hour: ${JWT_EXPIRE_HOUR:-24}
totp:
  encryption_key: "${TOTP_ENCRYPTION_KEY}"
  timezone: "${TZ:-Asia/Shanghai}"
default:
  user_concurrency: 5
  user_balance: 0
  api_key_prefix: "sk-"
  rate_multiplier: 1.0
  rate_limit:
    requests_per_minute: 60
    burst_size: 10
gateway:
  max_account_switches: ${GATEWAY_MAX_ACCOUNT_SWITCHES:-3}
  max_account_switches_gemini: ${GATEWAY_MAX_ACCOUNT_SWITCHES_GEMINI:-2}
  max_idle_conns: ${GATEWAY_MAX_IDLE_CONNS:-256}
  max_idle_conns_per_host: ${GATEWAY_MAX_IDLE_CONNS_PER_HOST:-32}
  max_conns_per_host: ${GATEWAY_MAX_CONNS_PER_HOST:-128}
  max_upstream_clients: ${GATEWAY_MAX_UPSTREAM_CLIENTS:-256}
  client_idle_ttl_seconds: ${GATEWAY_CLIENT_IDLE_TTL_SECONDS:-300}
  stream_keepalive_interval: ${GATEWAY_STREAM_KEEPALIVE_INTERVAL:-5}
  models_list_cache_ttl_seconds: ${GATEWAY_MODELS_LIST_CACHE_TTL_SECONDS:-30}
  openai_http2:
    enabled: ${OPENAI_HTTP2_ENABLED:-true}
    allow_proxy_fallback_to_http1: ${OPENAI_HTTP2_ALLOW_PROXY_FALLBACK_TO_HTTP1:-true}
  openai_ws:
    max_conns_per_account: ${OPENAI_WS_MAX_CONNS_PER_ACCOUNT:-24}
    min_idle_per_account: ${OPENAI_WS_MIN_IDLE_PER_ACCOUNT:-0}
    max_idle_per_account: ${OPENAI_WS_MAX_IDLE_PER_ACCOUNT:-4}
    event_flush_batch_size: ${OPENAI_WS_EVENT_FLUSH_BATCH_SIZE:-1}
    event_flush_interval_ms: ${OPENAI_WS_EVENT_FLUSH_INTERVAL_MS:-5}
usage_record:
  worker_count: ${USAGE_RECORD_WORKER_COUNT:-16}
  queue_size: ${USAGE_RECORD_QUEUE_SIZE:-4096}
  task_timeout_seconds: ${USAGE_RECORD_TASK_TIMEOUT_SECONDS:-15}
  overflow_policy: "${USAGE_RECORD_OVERFLOW_POLICY:-sync}"
  auto_scale_enabled: ${USAGE_RECORD_AUTO_SCALE_ENABLED:-false}
EOF
    printf 'installed_at=%s\n' "$(date -u +%Y-%m-%dT%H:%M:%SZ)" >"$DATA_DIR/.installed"
    chown sub2api:sub2api "$DATA_DIR/config.yaml" "$DATA_DIR/.installed"
    chmod 600 "$DATA_DIR/config.yaml"
    chmod 400 "$DATA_DIR/.installed"
}

maybe_skip_auto_setup() {
    [ "${FORCE_RUNTIME_CONFIG:-false}" = "true" ] && {
        write_runtime_config
        export AUTO_SETUP=false
        echo "Runtime config forced; skipping auto setup."
        return 0
    }
    [ "${SKIP_AUTO_SETUP_WHEN_MIGRATED:-true}" != "true" ] && return 0
    applied_count="$(migration_count | tr -d '[:space:]')"
    case "$applied_count" in ''|*[!0-9]*) applied_count=0 ;; esac
    required_count="${MIGRATION_BOOTSTRAP_MIN_COUNT:-178}"
    echo "Detected applied migrations: ${applied_count}/${required_count}"
    if [ "$applied_count" -ge "$required_count" ]; then
        write_runtime_config
        export AUTO_SETUP=false
        echo "External database is already bootstrapped; skipping auto setup."
    fi
}

sync_admin_password() {
    [ "${SYNC_ADMIN_PASSWORD:-true}" != "true" ] && return 0
    echo "Admin password sync loop started."
    for attempt in $(seq 1 15); do
        password_hash="$(python3 - <<'PY'
import bcrypt, os
print(bcrypt.hashpw(os.environ["ADMIN_PASSWORD"].encode("utf-8"), bcrypt.gensalt()).decode("utf-8"))
PY
        )"
        table_exists="$(PGPASSWORD="${DATABASE_PASSWORD:-}" psql -h "$DATABASE_HOST" -p "$DATABASE_PORT" -U "$DATABASE_USER" -d "$DATABASE_DBNAME" -tAc "select to_regclass('public.users') is not null;" 2>/tmp/admin-user-check.err || true)"
        table_exists="$(printf '%s' "$table_exists" | tr -d '[:space:]')"
        if [ "$table_exists" = "t" ]; then
            PGPASSWORD="${DATABASE_PASSWORD:-}" psql -h "$DATABASE_HOST" -p "$DATABASE_PORT" -U "$DATABASE_USER" -d "$DATABASE_DBNAME" -v ON_ERROR_STOP=1 \
                -c "update users set password_hash = '${password_hash}', role = 'admin', status = 'active', updated_at = now() where email = '${ADMIN_EMAIL}'; insert into users (email, password_hash, role, balance, concurrency, status, created_at, updated_at) select '${ADMIN_EMAIL}', '${password_hash}', 'admin', 0, 5, 'active', now(), now() where not exists (select 1 from users where email = '${ADMIN_EMAIL}');" >/tmp/admin-password-sync.out 2>/tmp/admin-password-sync.err
            echo "Admin password synchronized from secret."
            return 0
        fi
        sleep 2
    done
    echo "Admin password sync skipped: admin user not found before timeout." >&2
}

# ============================================================
# 11. 主流程
# ============================================================
maybe_skip_auto_setup
start_balance_cache_sync

attempt=1
max_attempts="${BOOTSTRAP_MAX_ATTEMPTS:-6}"
while :; do
    echo "Starting gateway on ${SERVER_HOST}:${SERVER_PORT} (attempt ${attempt}/${max_attempts})"
    su-exec sub2api /app/hub-gateway &
    app_pid="$!"
    trap 'kill -TERM "$app_pid" 2>/dev/null || true' INT TERM

    sync_admin_password &
    set +e
    wait "$app_pid"
    app_status="$?"
    set -e

    if [ "$app_status" -eq 0 ]; then
        exit 0
    fi

    if [ "$AUTO_SETUP" != "true" ]; then
        exit "$app_status"
    fi

    maybe_skip_auto_setup
    if [ "$AUTO_SETUP" != "true" ]; then
        attempt=$((attempt + 1))
        continue
    fi

    if [ "$attempt" -ge "$max_attempts" ]; then
        exit "$app_status"
    fi

    attempt=$((attempt + 1))
    echo "Gateway exited during bootstrap; retrying to continue migrations."
    sleep 2
done