v0.2.2

Wei-Shaw/sub2apiv0.2.2Sep 7, 2026by github-actions[bot]

AI Summary

Sub2API 0.2.2 introduces a Group Model Allowlist feature for precise model control and expands Simple Mode with account grouping capabilities.

Key Highlights

  • Group Model Allowlist allows administrators to configure specific model white lists per group, with the gateway intercepting requests for unauthorized models.
  • Simple Mode now supports account grouping, capability allowlists, and boundary controls for better resource management.
  • Reasoning intensity mapping can now be set to 'Reject' to prevent specific intensity levels from being used.
  • The admin backend now supports direct navigation to user usage statistics from the subscription management page.

Breaking Changes

  • Group Model Allowlist: Once enabled, models not present in the allowlist will be rejected by the gateway. Users must review and update group configurations after upgrading.

New Features

  • Group Model Allowlist: Configure available model white lists per group.
  • Reasoning Intensity Mapping Rejection: Set to 'Reject' to block specific intensity levels.
  • Simple Mode Group Management: Support for account grouping and boundary control.
  • Account Expiry Presets: Quick options for setting month or year expiration dates.
  • Subscription Usage Linkage: Direct jump to view user usage data in the admin panel.

Full Release Notes

> AI API Gateway Platform - 将 AI 订阅配额分发和管理

新增分组模型白名单功能,支持精确控制每个分组可用的模型范围;简单模式现已支持账号分组管理。

## 新增功能

- 分组模型白名单:支持为每个分组配置可用模型白名单,网关端点统一拦截
- 推理强度映射拒绝:分组的推理强度映射现在支持设置为"拒绝",禁止特定强度级别
- 简单模式分组管理:简单模式下支持基础的账号分组、能力白名单和分组边界管控
- 账号到期预设:新增月/年到期时间快捷选项
- 订阅用量关联:管理后台订阅页支持直接跳转查看用户用量

## 优化改进

- 模型广场定价列适配缓存价格和分时倍率定价的展示
- 流式失败时记录上游诊断信息,提升排障效率
- 连接测试改用账号实际可用模型,减少误报

## Bug 修复

- 修复 DeepSeek 高峰时段账号成本少计问题
- 修复 GLM-5.3/GLM-5.3-Flash 被 GLM-5 子串匹配抢走兜底价
- 修复 Gemini 3.7/3.8 Flash 思考层级定价不正确
- 修复 model-not-found 响应未触发 failover
- 修复 Codex 清单的 model_routing 和调度未生效
- 修复混合账号类型下默认模型列表丢失
- 修复 WebSocket 配额耗尽后后续 turn 无法恢复
- 修复 WebSocket HTTP 桥接连接状态未隔离
- 修复 WebSocket 透传 turn 槽位未正确回收
- 修复 Claude Code 版本伪装未生效导致 Fable 5.1 返回 400
- 修复 Claude 思考模式 block binding 缺失
- 修复 Codex allowed_tools 选择约束丢失
- 修复 Astra Ultra 工作流元数据和推理模式未保留
- 修复 GPT-6 Astra 系统指令选择错误
- 修复 Grok 媒体账号因不确定计费结果被排除
- 修复 Fable 额度错误范围过宽影响其他模型
- 修复 Antigravity 推理模型缺少 toolConfig 配置
- 修复消息缓存提升时 JSON 转义不正确
- 修复 Responses API 命名的独立输入被丢弃
- 修复数据库备份与迁移的 advisory lock 未串行化
- 修复支付履约与兑换码限流相互干扰
- 修复兑换码失败窗口从每小时改为固定 10 分钟
- 修复 Gemini 监控探测缺少 user 角色字段
- 修复 Codex 清单编辑响应式绑定和固定账号发现
- 修复用量页面 API Key 筛选器未加载全部 Key
- 修复引导页按钮文字阴影
- 修复 Turnstile 验证加载状态缺少占位

## 破坏性变更

- 分组模型白名单(Group Model Allowlist):启用后,未在白名单中的模型将被网关拒绝。升级后请检查分组配置,确认白名单设置符合预期。



---

## 📥 Installation

**Docker:**
```bash
# Docker Hub
docker pull weishaw/sub2api:0.2.2

# GitHub Container Registry
docker pull ghcr.io/wei-shaw/sub2api:0.2.2
```

**One-line install (Linux):**
```bash
curl -sSL https://raw.githubusercontent.com/Wei-Shaw/sub2api/main/deploy/install.sh | sudo bash
```

**Manual download:**
Download the appropriate archive for your platform from the assets below.

## 📚 Documentation

- [GitHub Repository](https://github.com/Wei-Shaw/sub2api)
- [Installation Guide](https://github.com/Wei-Shaw/sub2api/blob/main/deploy/README.md)