Skip to content

Releases

TokenRouter 0.2.3-dev.1

Pre-release

Choose a tag to compare

@github-actions github-actions released this 03 Oct 13:08
  • 上游 0.2.3(4878b58ac)23 个提交全部合入
  • fork: 第三方生图 / 创作台重试与历史 / 广场分时高峰展示价 / zcode 与 cline_pass 适配器

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.2.3-dev.1

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.2.1-1

TokenRouter 0.2.1-1 Pre-release
Pre-release

Choose a tag to compare

@github-actions github-actions released this 02 Oct 11:01

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

基于上游 TokenFlux/TokenRouter v0.2.1(d81377f19)+ fork 4 个提交:

  • feat(creative): 支持经 OpenAI 兼容协议接入的第三方生图模型(qwen-image/wan2.7-image/gemini-3.1-flash-lite-image 契约登记与执行)
  • feat(creative): 创作台失败重试与历史提示词复用
  • feat(marketplace): 广场展示价叠加分时与高峰倍率并标注时段
  • fix(creative): 明文 HTTP 下 crypto.randomUUID 缺失兜底

显式 0 价图片展示已由上游指针语义覆盖,原 fork 提交不再保留。


📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.2.1-1

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.292

Choose a tag to compare

@github-actions github-actions released this 23 Sep 21:17

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

  • 创作台服务端只保存 prompt 的 sha256,失败任务无法从服务端取回提示词,因此在创建成功时把提交参数快照留在当前浏览器(IndexedDB settings store,上限 50 条,超出按最旧清理;本地存储不可用时静默失败)
  • 历史记录:终态且非 succeeded 的任务(failed/cancelled/result_lost)展开后提供「重试」,先用本地快照还原输入区,再走与点击生成完全相同的画布采集与提交路径
  • 输入区新增「历史提示词」入口:按最近使用倒序、按提示词文本去重,点击直接填入提示词
  • 本机无快照或模型已下架时给出明确提示,不静默失败

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.292

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.291

Choose a tag to compare

@github-actions github-actions released this 23 Sep 20:29

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

  • 创作台对 qwen-image-2.0 / wan2.7-image / gemini-3.1-flash-lite-image 建单成功但执行必失败:creativePlatformImageModel 在账号/渠道映射后只认 gpt-image-*,目录能列出而执行器以 'creative mapped model X is not an image model' 拒绝
  • 该错误不可重试,直接 FailRun 且不写日志,只能从任务详情看到
  • 改用统一判定 isCreativeOpenAIImageModel(GPT Image 原生族 + 已登记第三方兼容生图模型),与目录候选、能力表、尺寸与请求体保持同一来源
  • 不含 image 关键字的普通文本模型仍被拒;新增回归测试覆盖放行与拒绝两侧

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.291

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.290

Choose a tag to compare

@github-actions github-actions released this 23 Sep 18:40

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

  • 原先 /v1/images/* 只放行 gpt-image-* 与 grok-imagine*,导致经 OpenAI 兼容协议接入的第三方生图模型(qwen-image-2.0、wan2.7-image、gemini-3.1-flash-lite-image)被 400 拒绝;创作台走执行器直连上游不受此限
  • 改为复用生图别名判定(isOpenAIImageGenerationModel 或名称含 image),只在端点校验层放宽
  • 不动共享的 isOpenAIImageGenerationModel:该函数被图片意图识别、账号图片限流、Codex 生图桥接与定价共 6 处依赖,放宽会误判文本模型
  • 不含 image 关键字的普通文本模型仍被拒;新增回归测试覆盖放行与拒绝两侧

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.290

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.289

Choose a tag to compare

@github-actions github-actions released this 23 Sep 17:50

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

  • 渠道新增「同步账号支持模型」:用该渠道下账号的已知支持模型并集覆盖渠道价卡;已有条目保留原价、新增模型显式填 0、并集外模型连同价卡移除并需确认;可选并发从上游刷新账号模型(失效账号忽略)
  • 账号「同步上游支持的模型」由只追加改为覆盖式替换,上游已下线模型不再残留;新增 apply 参数支持服务端直接覆盖写库(空结果不改并报错)
  • 创作台接入 OpenAI 兼容第三方生图模型:按模型名前缀登记请求契约(size 格式、是否显式索要 base64、编辑源图上限),新增同类模型只需登记一条
  • 已登记 gemini-3.1-flash-lite-image(size 传比例,默认 1:1)、qwen-image-2.0 与 wan2.7-image(size 传 WIDTHxHEIGHT,需显式请求 response_format=b64_json)
  • deploy 下 compose、脚本、示例与文档的镜像引用统一改为本仓库 ghcr.io/yegetables/tokenrouter
  • AGENTS.md 增加「渠道信息源」规范:渠道任务先查 .agents/channels/,禁止写死易变数据,获得新信息后回写维护

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.289

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.288

Choose a tag to compare

@github-actions github-actions released this 21 Sep 15:14

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

  • 上游在 SSE 事件中途断连(HTTP/2 client connection lost、连接被 reset 等)时,网关不再把"没有换行结尾的半截帧"透传给客户端,也不再把随后补发的合成错误帧粘进同一个 SSE 事件
  • 修复前:客户端收到的 data 是两个 JSON 挤在一个字段,严格客户端(如 opencode)报 Invalid ... stream event 并中断整轮对话
  • 修复后:客户端只收到完整事件 + 一个独立错误事件,能读到真实原因(如 Upstream response stream was interrupted)
  • 共享 SSE scanner 丢弃未以换行结束的残行并记录 droppedFragment;raw Chat Completions 路径把该信号补回截断判定,只收到残行的请求仍按上游截断处理,不会退化成 200 成功
  • 注入合成错误帧前补 SSE 事件边界空行(对齐上游 TokenFlux#1479 的做法),仅在本请求已写出响应体字节时补,避免流首多出空事件
  • 新增 6 个回归测试并做变异校验(去掉修复即失败,并复现线上畸形帧);同步 Project Doc 流式边界不变量

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.288

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.287

Choose a tag to compare

@github-actions github-actions released this 19 Sep 18:41

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

  • 新增「用量与缓存」表(用户端与管理端共用):模型 / 请求 / 输入 / 缓存 / 缓存率 / 输出 / 额度
  • 命中率口径统一为「缓存读取 ÷(输入 + 缓存创建 + 缓存读取)」,抽取 utils/cacheHitRate 复用
  • 上游不回报缓存写入时显示「不适用」,命中率仍按缓存读取计算,不再误报为 0%
  • 模型分布父行与按用户子行新增缓存率列;UserBreakdownItem 补缓存创建/读取拆分
  • Token 趋势图无流量时间点断线,不再画成 0% 造成「未命中」误读
  • 用量与缓存表跟随页面时间范围;移除全量查询路径
  • 实扣(站点余额单位)与标准/成本(USD 基准价)加口径说明 tooltip
  • 新增前端单测(命中率口径 / 不适用状态 / 趋势断线)与后端拆分字段测试

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.287

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.286

Choose a tag to compare

@github-actions github-actions released this 11 Sep 15:40

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

  • 新增 POST /api/v1/keys/:id/rotate:保留 Key ID、名称、分组、有效期、限额、IP 规则、模型映射、结算方式与全部用量/计费历史,仅替换鉴权凭据值
  • 仓储 CAS 更新,并发轮换返回冲突;成功后失效旧、新凭据的鉴权缓存,数据库触发器经 outbox 跨实例广播
  • 前端「更多」菜单新增「轮换密钥」并二次确认;审计动作 user.api_key.rotate
  • 同步 http_api 与 routing_and_billing 工程文档

📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.286

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation

TokenRouter 0.1.285

Choose a tag to compare

@github-actions github-actions released this 11 Sep 13:11

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


📥 Installation

Docker:

# GitHub Container Registry
docker pull ghcr.io/yegetables/tokenrouter:0.1.285

One-line install (Linux):

curl -sSL https://raw.githubusercontent.com/yegetables/TokenRouter/main/deploy/install.sh | sudo bash

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

📚 Documentation