refactor(数据库): 迁移持久层到 Ent 并清理 GORM

将仓储层/基础设施改为 Ent + 原生 SQL 执行路径，并移除 AutoMigrate 与 GORM 依赖。
重构内容包括：
- 仓储层改用 Ent/SQL（含 usage_log/account 等复杂查询），统一错误映射
- 基础设施与 setup 初始化切换为 Ent + SQL migrations
- 集成测试与 fixtures 迁移到 Ent 事务模型
- 清理遗留 GORM 模型/依赖，补充迁移与文档说明
- 增加根目录 Makefile 便于前后端编译

测试：
- go test -tags unit ./...
- go test -tags integration ./...

backend/migrations/001_init.sql

@@ -170,14 +170,3 @@ CREATE INDEX IF NOT EXISTS idx_usage_logs_account_id ON usage_logs(account_id);
 CREATE INDEX IF NOT EXISTS idx_usage_logs_model ON usage_logs(model);
 CREATE INDEX IF NOT EXISTS idx_usage_logs_created_at ON usage_logs(created_at);
 CREATE INDEX IF NOT EXISTS idx_usage_logs_user_created ON usage_logs(user_id, created_at);
-
--- 插入默认管理员用户
--- 密码: admin123 (bcrypt hash)
-INSERT INTO users (email, password_hash, role, balance, concurrency, status)
-VALUES ('admin@sub2api.com', '$2a$10$N9qo8uLOickgx2ZMRZoMye.IjJbDdJeCo0U2bBPJj9lS/5LqD.C.C', 'admin', 0, 10, 'active')
-ON CONFLICT (email) DO NOTHING;
-
--- 插入默认分组
-INSERT INTO groups (name, description, rate_multiplier, is_exclusive, status)
-VALUES ('default', '默认分组', 1.0, false, 'active')
-ON CONFLICT (name) DO NOTHING;

backend/migrations/005_schema_parity.sql

@@ -0,0 +1,42 @@
+-- Align SQL migrations with current GORM persistence models.
+-- This file is designed to be safe on both fresh installs and existing databases.
+
+-- users: add fields added after initial migration
+ALTER TABLE users ADD COLUMN IF NOT EXISTS username VARCHAR(100) NOT NULL DEFAULT '';
+ALTER TABLE users ADD COLUMN IF NOT EXISTS wechat VARCHAR(100) NOT NULL DEFAULT '';
+ALTER TABLE users ADD COLUMN IF NOT EXISTS notes TEXT NOT NULL DEFAULT '';
+
+-- api_keys: allow longer keys (GORM model uses size:128)
+ALTER TABLE api_keys ALTER COLUMN key TYPE VARCHAR(128);
+
+-- accounts: scheduling and rate-limit fields used by repository queries
+ALTER TABLE accounts ADD COLUMN IF NOT EXISTS schedulable BOOLEAN NOT NULL DEFAULT TRUE;
+ALTER TABLE accounts ADD COLUMN IF NOT EXISTS rate_limited_at TIMESTAMPTZ;
+ALTER TABLE accounts ADD COLUMN IF NOT EXISTS rate_limit_reset_at TIMESTAMPTZ;
+ALTER TABLE accounts ADD COLUMN IF NOT EXISTS overload_until TIMESTAMPTZ;
+ALTER TABLE accounts ADD COLUMN IF NOT EXISTS session_window_start TIMESTAMPTZ;
+ALTER TABLE accounts ADD COLUMN IF NOT EXISTS session_window_end TIMESTAMPTZ;
+ALTER TABLE accounts ADD COLUMN IF NOT EXISTS session_window_status VARCHAR(20);
+
+CREATE INDEX IF NOT EXISTS idx_accounts_schedulable ON accounts(schedulable);
+CREATE INDEX IF NOT EXISTS idx_accounts_rate_limited_at ON accounts(rate_limited_at);
+CREATE INDEX IF NOT EXISTS idx_accounts_rate_limit_reset_at ON accounts(rate_limit_reset_at);
+CREATE INDEX IF NOT EXISTS idx_accounts_overload_until ON accounts(overload_until);
+
+-- redeem_codes: subscription redeem fields
+ALTER TABLE redeem_codes ADD COLUMN IF NOT EXISTS group_id BIGINT REFERENCES groups(id) ON DELETE SET NULL;
+ALTER TABLE redeem_codes ADD COLUMN IF NOT EXISTS validity_days INT NOT NULL DEFAULT 30;
+CREATE INDEX IF NOT EXISTS idx_redeem_codes_group_id ON redeem_codes(group_id);
+
+-- usage_logs: billing type used by filters and stats
+ALTER TABLE usage_logs ADD COLUMN IF NOT EXISTS billing_type SMALLINT NOT NULL DEFAULT 0;
+CREATE INDEX IF NOT EXISTS idx_usage_logs_billing_type ON usage_logs(billing_type);
+
+-- settings: key-value store
+CREATE TABLE IF NOT EXISTS settings (
+    id          BIGSERIAL PRIMARY KEY,
+    key         VARCHAR(100) NOT NULL UNIQUE,
+    value       TEXT NOT NULL,
+    updated_at  TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+

backend/migrations/006_fix_invalid_subscription_expires_at.sql

@@ -0,0 +1,10 @@
+-- Fix legacy subscription records with invalid expires_at (year > 2099).
+DO $$
+BEGIN
+  IF to_regclass('public.user_subscriptions') IS NOT NULL THEN
+    UPDATE user_subscriptions
+    SET expires_at = TIMESTAMPTZ '2099-12-31 23:59:59+00'
+    WHERE expires_at > TIMESTAMPTZ '2099-12-31 23:59:59+00';
+  END IF;
+END $$;
+

backend/migrations/007_add_user_allowed_groups.sql

@@ -0,0 +1,20 @@
+-- Add user_allowed_groups join table to replace users.allowed_groups (BIGINT[]).
+-- Phase 1: create table + backfill from the legacy array column.
+
+CREATE TABLE IF NOT EXISTS user_allowed_groups (
+    user_id     BIGINT NOT NULL REFERENCES users(id) ON DELETE CASCADE,
+    group_id    BIGINT NOT NULL REFERENCES groups(id) ON DELETE CASCADE,
+    created_at  TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    PRIMARY KEY (user_id, group_id)
+);
+
+CREATE INDEX IF NOT EXISTS idx_user_allowed_groups_group_id ON user_allowed_groups(group_id);
+
+-- Backfill from the legacy users.allowed_groups array.
+INSERT INTO user_allowed_groups (user_id, group_id)
+SELECT u.id, x.group_id
+FROM users u
+CROSS JOIN LATERAL unnest(u.allowed_groups) AS x(group_id)
+JOIN groups g ON g.id = x.group_id
+WHERE u.allowed_groups IS NOT NULL
+ON CONFLICT DO NOTHING;

backend/migrations/migrations.go

@@ -0,0 +1,34 @@
+// Package migrations 包含嵌入的 SQL 数据库迁移文件。
+//
+// 该包使用 Go 1.16+ 的 embed 功能将 SQL 文件嵌入到编译后的二进制文件中。
+// 这种方式的优点：
+//   - 部署时无需额外的迁移文件
+//   - 迁移文件与代码版本一致
+//   - 便于版本控制和代码审查
+package migrations
+
+import "embed"
+
+// FS 包含本目录下所有嵌入的 SQL 迁移文件。
+//
+// 迁移命名规范：
+//   - 使用零填充的数字前缀确保正确的执行顺序
+//   - 格式：NNN_description.sql（如 001_init.sql, 002_add_users.sql）
+//   - 描述部分使用下划线分隔的小写单词
+//
+// 迁移文件要求：
+//   - 必须是幂等的（可重复执行而不产生错误）
+//   - 推荐使用 IF NOT EXISTS / IF EXISTS 语法
+//   - 一旦应用，不应修改已有的迁移文件（通过 checksum 校验）
+//
+// 示例迁移文件：
+//
+//	-- 001_init.sql
+//	CREATE TABLE IF NOT EXISTS users (
+//	    id BIGSERIAL PRIMARY KEY,
+//	    email VARCHAR(255) NOT NULL UNIQUE,
+//	    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+//	);
+//
+//go:embed *.sql
+var FS embed.FS