feat: add Google Gemini Pro support (#826)

* fest: Add Google Gemini Pro, fix #810

* fest: Add tooling to Gemini; Add OpenAI-like system prompt to Gemini

* refactor: removing unused if statement

* fest: Add dummy model message for system message in gemini model

* chore: update implementation

---------

Co-authored-by: JustSong <songquanpeng@foxmail.com>

.gitignore

@@ -4,4 +4,6 @@ upload
 *.exe
 *.db
 build
-*.db-journal
+*.db-journal
+logs
+data

README.en.md

@@ -60,7 +60,7 @@ _✨ Access all LLM through the standard OpenAI API format, easy to deploy & use
 1. Support for multiple large models:
    + [x] [OpenAI ChatGPT Series Models](https://platform.openai.com/docs/guides/gpt/chat-completions-api) (Supports [Azure OpenAI API](https://learn.microsoft.com/en-us/azure/ai-services/openai/reference))
    + [x] [Anthropic Claude Series Models](https://anthropic.com)
-   + [x] [Google PaLM2 Series Models](https://developers.generativeai.google)
+   + [x] [Google PaLM2 and Gemini Series Models](https://developers.generativeai.google)
    + [x] [Baidu Wenxin Yiyuan Series Models](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html)
    + [x] [Alibaba Tongyi Qianwen Series Models](https://help.aliyun.com/document_detail/2400395.html)
    + [x] [Zhipu ChatGLM Series Models](https://bigmodel.cn)
@@ -189,6 +189,8 @@ If you encounter a blank page after deployment, refer to [#97](https://github.co
 
 > Zeabur's servers are located overseas, automatically solving network issues, and the free quota is sufficient for personal usage.
 
+[![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/templates/7Q0KO3)
+
 1. First, fork the code.
 2. Go to [Zeabur](https://zeabur.com?referralCode=songquanpeng), log in, and enter the console.
 3. Create a new project. In Service -> Add Service, select Marketplace, and choose MySQL. Note down the connection parameters (username, password, address, and port).

README.ja.md

@@ -60,7 +60,7 @@ _✨ 標準的な OpenAI API フォーマットを通じてすべての LLM に
 1. 複数の大型モデルをサポート:
    + [x] [OpenAI ChatGPT シリーズモデル](https://platform.openai.com/docs/guides/gpt/chat-completions-api) ([Azure OpenAI API](https://learn.microsoft.com/en-us/azure/ai-services/openai/reference) をサポート)
    + [x] [Anthropic Claude シリーズモデル](https://anthropic.com)
-   + [x] [Google PaLM2 シリーズモデル](https://developers.generativeai.google)
+   + [x] [Google PaLM2/Gemini シリーズモデル](https://developers.generativeai.google)
    + [x] [Baidu Wenxin Yiyuan シリーズモデル](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html)
    + [x] [Alibaba Tongyi Qianwen シリーズモデル](https://help.aliyun.com/document_detail/2400395.html)
    + [x] [Zhipu ChatGLM シリーズモデル](https://bigmodel.cn)
@@ -190,6 +190,8 @@ Please refer to the [environment variables](#environment-variables) section for
 
 > Zeabur のサーバーは海外にあるため、ネットワークの問題は自動的に解決されます。
 
+[![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/templates/7Q0KO3)
+
 1. まず、コードをフォークする。
 2. [Zeabur](https://zeabur.com?referralCode=songquanpeng) にアクセスしてログインし、コンソールに入る。
 3. 新しいプロジェクトを作成します。Service -> Add ServiceでMarketplace を選択し、MySQL を選択する。接続パラメータ（ユーザー名、パスワード、アドレス、ポート）をメモします。

README.md

@@ -51,30 +51,34 @@ _✨ 通过标准的 OpenAI API 格式访问所有的大模型，开箱即用
   <a href="https://iamazing.cn/page/reward">赞赏支持</a>
 </p>
 
-> **Note**
+> [!NOTE]
 > 本项目为开源项目，使用者必须在遵循 OpenAI 的[使用条款](https://openai.com/policies/terms-of-use)以及**法律法规**的情况下使用，不得用于非法用途。
 > 
 > 根据[《生成式人工智能服务管理暂行办法》](http://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm)的要求，请勿对中国地区公众提供一切未经备案的生成式人工智能服务。
 
-> **Warning**
+> [!WARNING]
 > 使用 Docker 拉取的最新镜像可能是 `alpha` 版本，如果追求稳定性请手动指定版本。
 
+> [!WARNING]
+> 使用 root 用户初次登录系统后，务必修改默认密码 `123456`！
+
 ## 功能
 1. 支持多种大模型：
    + [x] [OpenAI ChatGPT 系列模型](https://platform.openai.com/docs/guides/gpt/chat-completions-api)（支持 [Azure OpenAI API](https://learn.microsoft.com/en-us/azure/ai-services/openai/reference)）
    + [x] [Anthropic Claude 系列模型](https://anthropic.com)
-   + [x] [Google PaLM2 系列模型](https://developers.generativeai.google)
+   + [x] [Google PaLM2/Gemini 系列模型](https://developers.generativeai.google)
    + [x] [百度文心一言系列模型](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html)
    + [x] [阿里通义千问系列模型](https://help.aliyun.com/document_detail/2400395.html)
    + [x] [讯飞星火认知大模型](https://www.xfyun.cn/doc/spark/Web.html)
    + [x] [智谱 ChatGLM 系列模型](https://bigmodel.cn)
    + [x] [360 智脑](https://ai.360.cn)
+   + [x] [腾讯混元大模型](https://cloud.tencent.com/document/product/1729)
 2. 支持配置镜像以及众多第三方代理服务：
    + [x] [OpenAI-SB](https://openai-sb.com)
+   + [x] [CloseAI](https://referer.shadowai.xyz/r/2412)
    + [x] [API2D](https://api2d.com/r/197971)
    + [x] [OhMyGPT](https://aigptx.top?aff=uFpUl2Kf)
    + [x] [AI Proxy](https://aiproxy.io/?i=OneAPI) （邀请码：`OneAPI`）
-   + [x] [CloseAI](https://console.closeai-asia.com/r/2412)
    + [x] 自定义渠道：例如各种未收录的第三方代理服务
 3. 支持通过**负载均衡**的方式访问多个渠道。
 4. 支持 **stream 模式**，可以通过流式传输实现打字机效果。
@@ -88,26 +92,35 @@ _✨ 通过标准的 OpenAI API 格式访问所有的大模型，开箱即用
 12. 支持**用户邀请奖励**。
 13. 支持以美元为单位显示额度。
 14. 支持发布公告，设置充值链接，设置新用户初始额度。
-15. 支持模型映射，重定向用户的请求模型。
+15. 支持模型映射，重定向用户的请求模型，如无必要请不要设置，设置之后会导致请求体被重新构造而非直接透传，会导致部分还未正式支持的字段无法传递成功。
 16. 支持失败自动重试。
 17. 支持绘图接口。
-18. 支持丰富的**自定义**设置，
+18. 支持 [Cloudflare AI Gateway](https://developers.cloudflare.com/ai-gateway/providers/openai/)，渠道设置的代理部分填写 `https://gateway.ai.cloudflare.com/v1/ACCOUNT_TAG/GATEWAY/openai` 即可。
+19. 支持丰富的**自定义**设置，
     1. 支持自定义系统名称，logo 以及页脚。
     2. 支持自定义首页和关于页面，可以选择使用 HTML & Markdown 代码进行自定义，或者使用一个单独的网页通过 iframe 嵌入。
-19. 支持通过系统访问令牌访问管理 API。
-20. 支持 Cloudflare Turnstile 用户校验。
-21. 支持用户管理，支持**多种用户登录注册方式**：
+20. 支持通过系统访问令牌访问管理 API（bearer token，用以替代 cookie，你可以自行抓包来查看 API 的用法）。
+21. 支持 Cloudflare Turnstile 用户校验。
+22. 支持用户管理，支持**多种用户登录注册方式**：
     + 邮箱登录注册（支持注册邮箱白名单）以及通过邮箱进行密码重置。
     + [GitHub 开放授权](https://github.com/settings/applications/new)。
     + 微信公众号授权（需要额外部署 [WeChat Server](https://github.com/songquanpeng/wechat-server)）。
 
 ## 部署
 ### 基于 Docker 进行部署
-部署命令：`docker run --name one-api -d --restart always -p 3000:3000 -e TZ=Asia/Shanghai -v /home/ubuntu/data/one-api:/data justsong/one-api`
+```shell
+# 使用 SQLite 的部署命令：
+docker run --name one-api -d --restart always -p 3000:3000 -e TZ=Asia/Shanghai -v /home/ubuntu/data/one-api:/data justsong/one-api
+# 使用 MySQL 的部署命令，在上面的基础上添加 `-e SQL_DSN="root:123456@tcp(localhost:3306)/oneapi"`，请自行修改数据库连接参数，不清楚如何修改请参见下面环境变量一节。
+# 例如：
+docker run --name one-api -d --restart always -p 3000:3000 -e SQL_DSN="root:123456@tcp(localhost:3306)/oneapi" -e TZ=Asia/Shanghai -v /home/ubuntu/data/one-api:/data justsong/one-api
+```
 
 其中，`-p 3000:3000` 中的第一个 `3000` 是宿主机的端口，可以根据需要进行修改。
 
-数据将会保存在宿主机的 `/home/ubuntu/data/one-api` 目录，请确保该目录存在且具有写入权限，或者更改为合适的目录。
+数据和日志将会保存在宿主机的 `/home/ubuntu/data/one-api` 目录，请确保该目录存在且具有写入权限，或者更改为合适的目录。
+
+如果启动失败，请添加 `--privileged=true`，具体参考 https://github.com/songquanpeng/one-api/issues/482 。
 
 如果上面的镜像无法拉取，可以尝试使用 GitHub 的 Docker 镜像，将上面的 `justsong/one-api` 替换为 `ghcr.io/songquanpeng/one-api` 即可。
 
@@ -147,6 +160,19 @@ sudo service nginx restart
 
 初始账号用户名为 `root`，密码为 `123456`。
 
+
+### 基于 Docker Compose 进行部署
+
+> 仅启动方式不同，参数设置不变，请参考基于 Docker 部署部分
+
+```shell
+# 目前支持 MySQL 启动，数据存储在 ./data/mysql 文件夹内
+docker-compose up -d
+
+# 查看部署状态
+docker-compose ps
+```
+
 ### 手动部署
 1. 从 [GitHub Releases](https://github.com/songquanpeng/one-api/releases/latest) 下载可执行文件或者从源码编译：
    ```shell
@@ -209,6 +235,13 @@ docker run --name chatgpt-web -d -p 3002:3002 -e OPENAI_API_BASE_URL=https://ope
 
 注意修改端口号、`OPENAI_API_BASE_URL` 和 `OPENAI_API_KEY`。
 
+#### QChatGPT - QQ机器人
+项目主页：https://github.com/RockChinQ/QChatGPT
+
+根据文档完成部署后，在`config.py`设置配置项`openai_config`的`reverse_proxy`为 One API 后端地址，设置`api_key`为 One API 生成的key，并在配置项`completion_api_params`的`model`参数设置为 One API 支持的模型名称。
+
+可安装 [Switcher 插件](https://github.com/RockChinQ/Switcher)在运行时切换所使用的模型。
+
 ### 部署到第三方平台
 <details>
 <summary><strong>部署到 Sealos </strong></summary>
@@ -227,7 +260,9 @@ docker run --name chatgpt-web -d -p 3002:3002 -e OPENAI_API_BASE_URL=https://ope
 <summary><strong>部署到 Zeabur</strong></summary>
 <div>
 
-> Zeabur 的服务器在国外，自动解决了网络的问题，同时免费的额度也足够个人使用。
+> Zeabur 的服务器在国外，自动解决了网络的问题，同时免费的额度也足够个人使用
+
+[![Deploy on Zeabur](https://zeabur.com/button.svg)](https://zeabur.com/templates/7Q0KO3)
 
 1. 首先 fork 一份代码。
 2. 进入 [Zeabur](https://zeabur.com?referralCode=songquanpeng)，登录，进入控制台。
@@ -242,6 +277,17 @@ docker run --name chatgpt-web -d -p 3002:3002 -e OPENAI_API_BASE_URL=https://ope
 </div>
 </details>
 
+<details>
+<summary><strong>部署到 Render</strong></summary>
+<div>
+
+> Render 提供免费额度，绑卡后可以进一步提升额度
+
+Render 可以直接部署 docker 镜像，不需要 fork 仓库：https://dashboard.render.com
+
+</div>
+</details>
+
 ## 配置
 系统本身开箱即用。
 
@@ -260,13 +306,20 @@ docker run --name chatgpt-web -d -p 3002:3002 -e OPENAI_API_BASE_URL=https://ope
 
 注意，具体的 API Base 的格式取决于你所使用的客户端。
 
+例如对于 OpenAI 的官方库：
+```bash
+OPENAI_API_KEY="sk-xxxxxx"
+OPENAI_API_BASE="https://<HOST>:<PORT>/v1" 
+```
+
 ```mermaid
 graph LR
     A(用户)
-    A --->|请求| B(One API)
+    A --->|使用 One API 分发的 key 进行请求| B(One API)
     B -->|中继请求| C(OpenAI)
     B -->|中继请求| D(Azure)
-    B -->|中继请求| E(其他下游渠道)
+    B -->|中继请求| E(其他 OpenAI API 格式下游渠道)
+    B -->|中继并修改请求体和返回体| F(非 OpenAI API 格式下游渠道)
 ```
 
 可以通过在令牌后面添加渠道 ID 的方式指定使用哪一个渠道处理本次请求，例如：`Authorization: Bearer ONE_API_KEY-CHANNEL_ID`。
@@ -275,8 +328,9 @@ graph LR
 不加的话将会使用负载均衡的方式使用多个渠道。
 
 ### 环境变量
-1. `REDIS_CONN_STRING`：设置之后将使用 Redis 作为请求频率限制的存储，而非使用内存存储。
+1. `REDIS_CONN_STRING`：设置之后将使用 Redis 作为缓存使用。
    + 例子：`REDIS_CONN_STRING=redis://default:redispw@localhost:49153`
+   + 如果数据库访问延迟很低，没有必要启用 Redis，启用后反而会出现数据滞后的问题。
 2. `SESSION_SECRET`：设置之后将使用固定的会话密钥，这样系统重新启动后已登录用户的 cookie 将依旧有效。
    + 例子：`SESSION_SECRET=random_string`
 3. `SQL_DSN`：设置之后将使用指定数据库而非 SQLite，请使用 MySQL 或 PostgreSQL。
@@ -293,21 +347,35 @@ graph LR
      + `SQL_CONN_MAX_LIFETIME`：连接的最大生命周期，默认为 `60`，单位分钟。
 4. `FRONTEND_BASE_URL`：设置之后将重定向页面请求到指定的地址，仅限从服务器设置。
    + 例子：`FRONTEND_BASE_URL=https://openai.justsong.cn`
-5. `SYNC_FREQUENCY`：设置之后将定期与数据库同步配置，单位为秒，未设置则不进行同步。
+5. `MEMORY_CACHE_ENABLED`：启用内存缓存，会导致用户额度的更新存在一定的延迟，可选值为 `true` 和 `false`，未设置则默认为 `false`。
+   + 例子：`MEMORY_CACHE_ENABLED=true`
+6. `SYNC_FREQUENCY`：在启用缓存的情况下与数据库同步配置的频率，单位为秒，默认为 `600` 秒。
    + 例子：`SYNC_FREQUENCY=60`
-6. `NODE_TYPE`：设置之后将指定节点类型，可选值为 `master` 和 `slave`，未设置则默认为 `master`。
+7. `NODE_TYPE`：设置之后将指定节点类型，可选值为 `master` 和 `slave`，未设置则默认为 `master`。
    + 例子：`NODE_TYPE=slave`
-7. `CHANNEL_UPDATE_FREQUENCY`：设置之后将定期更新渠道余额，单位为分钟，未设置则不进行更新。
+8. `CHANNEL_UPDATE_FREQUENCY`：设置之后将定期更新渠道余额，单位为分钟，未设置则不进行更新。
    + 例子：`CHANNEL_UPDATE_FREQUENCY=1440`
-8. `CHANNEL_TEST_FREQUENCY`：设置之后将定期检查渠道，单位为分钟，未设置则不进行检查。
+9. `CHANNEL_TEST_FREQUENCY`：设置之后将定期检查渠道，单位为分钟，未设置则不进行检查。
    + 例子：`CHANNEL_TEST_FREQUENCY=1440`
-9. `POLLING_INTERVAL`：批量更新渠道余额以及测试可用性时的请求间隔，单位为秒，默认无间隔。
-   + 例子：`POLLING_INTERVAL=5`
+10. `POLLING_INTERVAL`：批量更新渠道余额以及测试可用性时的请求间隔，单位为秒，默认无间隔。
+    + 例子：`POLLING_INTERVAL=5`
+11. `BATCH_UPDATE_ENABLED`：启用数据库批量更新聚合，会导致用户额度的更新存在一定的延迟可选值为 `true` 和 `false`，未设置则默认为 `false`。
+    + 例子：`BATCH_UPDATE_ENABLED=true`
+    + 如果你遇到了数据库连接数过多的问题，可以尝试启用该选项。
+12. `BATCH_UPDATE_INTERVAL=5`：批量更新聚合的时间间隔，单位为秒，默认为 `5`。
+    + 例子：`BATCH_UPDATE_INTERVAL=5`
+13. 请求频率限制：
+    + `GLOBAL_API_RATE_LIMIT`：全局 API 速率限制（除中继请求外），单 ip 三分钟内的最大请求数，默认为 `180`。
+    + `GLOBAL_WEB_RATE_LIMIT`：全局 Web 速率限制，单 ip 三分钟内的最大请求数，默认为 `60`。
+14. 编码器缓存设置：
+    + `TIKTOKEN_CACHE_DIR`：默认程序启动时会联网下载一些通用的词元的编码，如：`gpt-3.5-turbo`，在一些网络环境不稳定，或者离线情况，可能会导致启动有问题，可以配置此目录缓存数据，可迁移到离线环境。
+    + `DATA_GYM_CACHE_DIR`：目前该配置作用与 `TIKTOKEN_CACHE_DIR` 一致，但是优先级没有它高。
+15. `RELAY_TIMEOUT`：中继超时设置，单位为秒，默认不设置超时时间。
 
 ### 命令行参数
 1. `--port <port_number>`: 指定服务器监听的端口号，默认为 `3000`。
    + 例子：`--port 3000`
-2. `--log-dir <log_dir>`: 指定日志文件夹，如果没有设置，日志将不会被保存。
+2. `--log-dir <log_dir>`: 指定日志文件夹，如果没有设置，默认保存至工作目录的 `logs` 文件夹下。
    + 例子：`--log-dir ./logs`
 3. `--version`: 打印系统版本号并退出。
 4. `--help`: 查看命令的使用帮助和参数说明。
@@ -339,8 +407,15 @@ https://openai.justsong.cn
 5. ChatGPT Next Web 报错：`Failed to fetch`
    + 部署的时候不要设置 `BASE_URL`。
    + 检查你的接口地址和 API Key 有没有填对。
+   + 检查是否启用了 HTTPS，浏览器会拦截 HTTPS 域名下的 HTTP 请求。
 6. 报错：`当前分组负载已饱和，请稍后再试`
    + 上游通道 429 了。
+7. 升级之后我的数据会丢失吗？
+   + 如果使用 MySQL，不会。
+   + 如果使用 SQLite，需要按照我所给的部署命令挂载 volume 持久化 one-api.db 数据库文件，否则容器重启后数据会丢失。
+8. 升级之前数据库需要做变更吗？
+   + 一般情况下不需要，系统将在初始化的时候自动调整。
+   + 如果需要的话，我会在更新日志中说明，并给出脚本。
 
 ## 相关项目
 * [FastGPT](https://github.com/labring/FastGPT): 基于 LLM 大语言模型的知识库问答系统
@@ -352,4 +427,4 @@ https://openai.justsong.cn
 
 同样适用于基于本项目的二开项目。
 
-依据 MIT 协议，使用者需自行承担使用本项目的风险与责任，本开源项目开发者与此无关。
+依据 MIT 协议，使用者需自行承担使用本项目的风险与责任，本开源项目开发者与此无关。

common/constants.go

@@ -21,12 +21,9 @@ var QuotaPerUnit = 500 * 1000.0 // $0.002 / 1K tokens
 var DisplayInCurrencyEnabled = true
 var DisplayTokenStatEnabled = true
 
-var UsingSQLite = false
-
 // Any options with "Secret", "Token" in its key won't be return by GetOptions
 
 var SessionSecret = uuid.New().String()
-var SQLitePath = "one-api.db"
 
 var OptionMap map[string]string
 var OptionMapRWMutex sync.RWMutex
@@ -56,6 +53,7 @@ var EmailDomainWhitelist = []string{
 }
 
 var DebugEnabled = os.Getenv("DEBUG") == "true"
+var MemoryCacheEnabled = os.Getenv("MEMORY_CACHE_ENABLED") == "true"
 
 var LogConsumeEnabled = true
 
@@ -80,6 +78,7 @@ var QuotaForInviter = 0
 var QuotaForInvitee = 0
 var ChannelDisableThreshold = 5.0
 var AutomaticDisableChannelEnabled = false
+var AutomaticEnableChannelEnabled = false
 var QuotaRemindThreshold = 1000
 var PreConsumedQuota = 500
 var ApproximateTokenEnabled = false
@@ -92,7 +91,16 @@ var IsMasterNode = os.Getenv("NODE_TYPE") != "slave"
 var requestInterval, _ = strconv.Atoi(os.Getenv("POLLING_INTERVAL"))
 var RequestInterval = time.Duration(requestInterval) * time.Second
 
-var SyncFrequency = 10 * 60 // unit is second, will be overwritten by SYNC_FREQUENCY
+var SyncFrequency = GetOrDefault("SYNC_FREQUENCY", 10*60) // unit is second
+
+var BatchUpdateEnabled = false
+var BatchUpdateInterval = GetOrDefault("BATCH_UPDATE_INTERVAL", 5)
+
+var RelayTimeout = GetOrDefault("RELAY_TIMEOUT", 0) // unit is second
+
+const (
+	RequestIdKey = "X-Oneapi-Request-Id"
+)
 
 const (
 	RoleGuestUser  = 0
@@ -111,10 +119,10 @@ var (
 // All duration's unit is seconds
 // Shouldn't larger then RateLimitKeyExpirationDuration
 var (
-	GlobalApiRateLimitNum            = 180
+	GlobalApiRateLimitNum            = GetOrDefault("GLOBAL_API_RATE_LIMIT", 180)
 	GlobalApiRateLimitDuration int64 = 3 * 60
 
-	GlobalWebRateLimitNum            = 60
+	GlobalWebRateLimitNum            = GetOrDefault("GLOBAL_WEB_RATE_LIMIT", 60)
 	GlobalWebRateLimitDuration int64 = 3 * 60
 
 	UploadRateLimitNum            = 10
@@ -148,53 +156,64 @@ const (
 )
 
 const (
-	ChannelStatusUnknown  = 0
-	ChannelStatusEnabled  = 1 // don't use 0, 0 is the default value!
-	ChannelStatusDisabled = 2 // also don't use 0
+	ChannelStatusUnknown          = 0
+	ChannelStatusEnabled          = 1 // don't use 0, 0 is the default value!
+	ChannelStatusManuallyDisabled = 2 // also don't use 0
+	ChannelStatusAutoDisabled     = 3
 )
 
 const (
-	ChannelTypeUnknown   = 0
-	ChannelTypeOpenAI    = 1
-	ChannelTypeAPI2D     = 2
-	ChannelTypeAzure     = 3
-	ChannelTypeCloseAI   = 4
-	ChannelTypeOpenAISB  = 5
-	ChannelTypeOpenAIMax = 6
-	ChannelTypeOhMyGPT   = 7
-	ChannelTypeCustom    = 8
-	ChannelTypeAILS      = 9
-	ChannelTypeAIProxy   = 10
-	ChannelTypePaLM      = 11
-	ChannelTypeAPI2GPT   = 12
-	ChannelTypeAIGC2D    = 13
-	ChannelTypeAnthropic = 14
-	ChannelTypeBaidu     = 15
-	ChannelTypeZhipu     = 16
-	ChannelTypeAli       = 17
-	ChannelTypeXunfei    = 18
-	ChannelType360       = 19
+	ChannelTypeUnknown        = 0
+	ChannelTypeOpenAI         = 1
+	ChannelTypeAPI2D          = 2
+	ChannelTypeAzure          = 3
+	ChannelTypeCloseAI        = 4
+	ChannelTypeOpenAISB       = 5
+	ChannelTypeOpenAIMax      = 6
+	ChannelTypeOhMyGPT        = 7
+	ChannelTypeCustom         = 8
+	ChannelTypeAILS           = 9
+	ChannelTypeAIProxy        = 10
+	ChannelTypePaLM           = 11
+	ChannelTypeAPI2GPT        = 12
+	ChannelTypeAIGC2D         = 13
+	ChannelTypeAnthropic      = 14
+	ChannelTypeBaidu          = 15
+	ChannelTypeZhipu          = 16
+	ChannelTypeAli            = 17
+	ChannelTypeXunfei         = 18
+	ChannelType360            = 19
+	ChannelTypeOpenRouter     = 20
+	ChannelTypeAIProxyLibrary = 21
+	ChannelTypeFastGPT        = 22
+	ChannelTypeTencent        = 23
+	ChannelTypeGemini         = 24
 )
 
 var ChannelBaseURLs = []string{
-	"",                               // 0
-	"https://api.openai.com",         // 1
-	"https://oa.api2d.net",           // 2
-	"",                               // 3
-	"https://api.closeai-proxy.xyz",  // 4
-	"https://api.openai-sb.com",      // 5
-	"https://api.openaimax.com",      // 6
-	"https://api.ohmygpt.com",        // 7
-	"",                               // 8
-	"https://api.caipacity.com",      // 9
-	"https://api.aiproxy.io",         // 10
-	"",                               // 11
-	"https://api.api2gpt.com",        // 12
-	"https://api.aigc2d.com",         // 13
-	"https://api.anthropic.com",      // 14
-	"https://aip.baidubce.com",       // 15
-	"https://open.bigmodel.cn",       // 16
-	"https://dashscope.aliyuncs.com", // 17
-	"",                               // 18
-	"https://ai.360.cn",              // 19
+	"",                                  // 0
+	"https://api.openai.com",            // 1
+	"https://oa.api2d.net",              // 2
+	"",                                  // 3
+	"https://api.closeai-proxy.xyz",     // 4
+	"https://api.openai-sb.com",         // 5
+	"https://api.openaimax.com",         // 6
+	"https://api.ohmygpt.com",           // 7
+	"",                                  // 8
+	"https://api.caipacity.com",         // 9
+	"https://api.aiproxy.io",            // 10
+	"",                                  // 11
+	"https://api.api2gpt.com",           // 12
+	"https://api.aigc2d.com",            // 13
+	"https://api.anthropic.com",         // 14
+	"https://aip.baidubce.com",          // 15
+	"https://open.bigmodel.cn",          // 16
+	"https://dashscope.aliyuncs.com",    // 17
+	"",                                  // 18
+	"https://ai.360.cn",                 // 19
+	"https://openrouter.ai/api",         // 20
+	"https://api.aiproxy.io",            // 21
+	"https://fastgpt.run/api/openapi",   // 22
+	"https://hunyuan.cloud.tencent.com", //23
+	"",                                  //24
 }

common/database.go

@@ -0,0 +1,6 @@
+package common
+
+var UsingSQLite = false
+var UsingPostgreSQL = false
+
+var SQLitePath = "one-api.db"

common/email.go

@@ -1,11 +1,13 @@
 package common
 
 import (
+	"crypto/rand"
 	"crypto/tls"
 	"encoding/base64"
 	"fmt"
 	"net/smtp"
 	"strings"
+	"time"
 )
 
 func SendEmail(subject string, receiver string, content string) error {
@@ -13,15 +15,32 @@ func SendEmail(subject string, receiver string, content string) error {
 		SMTPFrom = SMTPAccount
 	}
 	encodedSubject := fmt.Sprintf("=?UTF-8?B?%s?=", base64.StdEncoding.EncodeToString([]byte(subject)))
+
+	// Extract domain from SMTPFrom
+	parts := strings.Split(SMTPFrom, "@")
+	var domain string
+	if len(parts) > 1 {
+		domain = parts[1]
+	}
+	// Generate a unique Message-ID
+	buf := make([]byte, 16)
+	_, err := rand.Read(buf)
+	if err != nil {
+		return err
+	}
+	messageId := fmt.Sprintf("<%x@%s>", buf, domain)
+
 	mail := []byte(fmt.Sprintf("To: %s\r\n"+
 		"From: %s<%s>\r\n"+
 		"Subject: %s\r\n"+
+		"Message-ID: %s\r\n"+ // add Message-ID header to avoid being treated as spam, RFC 5322
+		"Date: %s\r\n"+
 		"Content-Type: text/html; charset=UTF-8\r\n\r\n%s\r\n",
-		receiver, SystemName, SMTPFrom, encodedSubject, content))
+		receiver, SystemName, SMTPFrom, encodedSubject, messageId, time.Now().Format(time.RFC1123Z), content))
 	auth := smtp.PlainAuth("", SMTPAccount, SMTPToken, SMTPServer)
 	addr := fmt.Sprintf("%s:%d", SMTPServer, SMTPPort)
 	to := strings.Split(receiver, ";")
-	var err error
+
 	if SMTPPort == 465 {
 		tlsConfig := &tls.Config{
 			InsecureSkipVerify: true,

common/gin.go

@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"github.com/gin-gonic/gin"
 	"io"
+	"strings"
 )
 
 func UnmarshalBodyReusable(c *gin.Context, v any) error {
@@ -16,7 +17,13 @@ func UnmarshalBodyReusable(c *gin.Context, v any) error {
 	if err != nil {
 		return err
 	}
-	err = json.Unmarshal(requestBody, &v)
+	contentType := c.Request.Header.Get("Content-Type")
+	if strings.HasPrefix(contentType, "application/json") {
+		err = json.Unmarshal(requestBody, &v)
+	} else {
+		// skip for now
+		// TODO: someday non json request have variant model, we will need to implementation this
+	}
 	if err != nil {
 		return err
 	}

common/image/image.go

@@ -0,0 +1,47 @@
+package image
+
+import (
+	"image"
+	_ "image/gif"
+	_ "image/jpeg"
+	_ "image/png"
+	"net/http"
+	"regexp"
+	"strings"
+
+	_ "golang.org/x/image/webp"
+)
+
+func GetImageSizeFromUrl(url string) (width int, height int, err error) {
+	resp, err := http.Get(url)
+	if err != nil {
+		return
+	}
+	defer resp.Body.Close()
+	img, _, err := image.DecodeConfig(resp.Body)
+	if err != nil {
+		return
+	}
+	return img.Width, img.Height, nil
+}
+
+var (
+	reg = regexp.MustCompile(`data:image/([^;]+);base64,`)
+)
+
+func GetImageSizeFromBase64(encoded string) (width int, height int, err error) {
+	encoded = strings.TrimPrefix(encoded, "data:image/png;base64,")
+	base64 := strings.NewReader(reg.ReplaceAllString(encoded, ""))
+	img, _, err := image.DecodeConfig(base64)
+	if err != nil {
+		return
+	}
+	return img.Width, img.Height, nil
+}
+
+func GetImageSize(image string) (width int, height int, err error) {
+	if strings.HasPrefix(image, "data:image/") {
+		return GetImageSizeFromBase64(image)
+	}
+	return GetImageSizeFromUrl(image)
+}

common/image/image_test.go

@@ -0,0 +1,154 @@
+package image_test
+
+import (
+	"encoding/base64"
+	"image"
+	_ "image/gif"
+	_ "image/jpeg"
+	_ "image/png"
+	"io"
+	"net/http"
+	"strconv"
+	"strings"
+	"testing"
+
+	img "one-api/common/image"
+
+	"github.com/stretchr/testify/assert"
+	_ "golang.org/x/image/webp"
+)
+
+type CountingReader struct {
+	reader    io.Reader
+	BytesRead int
+}
+
+func (r *CountingReader) Read(p []byte) (n int, err error) {
+	n, err = r.reader.Read(p)
+	r.BytesRead += n
+	return n, err
+}
+
+var (
+	cases = []struct {
+		url    string
+		format string
+		width  int
+		height int
+	}{
+		{"https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg", "jpeg", 2560, 1669},
+		{"https://upload.wikimedia.org/wikipedia/commons/9/97/Basshunter_live_performances.png", "png", 4500, 2592},
+		{"https://upload.wikimedia.org/wikipedia/commons/c/c6/TO_THE_ONE_SOMETHINGNESS.webp", "webp", 984, 985},
+		{"https://upload.wikimedia.org/wikipedia/commons/d/d0/01_Das_Sandberg-Modell.gif", "gif", 1917, 1533},
+		{"https://upload.wikimedia.org/wikipedia/commons/6/62/102Cervus.jpg", "jpeg", 270, 230},
+	}
+)
+
+func TestDecode(t *testing.T) {
+	// Bytes read: varies sometimes
+	// jpeg: 1063892
+	// png: 294462
+	// webp: 99529
+	// gif: 956153
+	// jpeg#01: 32805
+	for _, c := range cases {
+		t.Run("Decode:"+c.format, func(t *testing.T) {
+			resp, err := http.Get(c.url)
+			assert.NoError(t, err)
+			defer resp.Body.Close()
+			reader := &CountingReader{reader: resp.Body}
+			img, format, err := image.Decode(reader)
+			assert.NoError(t, err)
+			size := img.Bounds().Size()
+			assert.Equal(t, c.format, format)
+			assert.Equal(t, c.width, size.X)
+			assert.Equal(t, c.height, size.Y)
+			t.Logf("Bytes read: %d", reader.BytesRead)
+		})
+	}
+
+	// Bytes read:
+	// jpeg: 4096
+	// png: 4096
+	// webp: 4096
+	// gif: 4096
+	// jpeg#01: 4096
+	for _, c := range cases {
+		t.Run("DecodeConfig:"+c.format, func(t *testing.T) {
+			resp, err := http.Get(c.url)
+			assert.NoError(t, err)
+			defer resp.Body.Close()
+			reader := &CountingReader{reader: resp.Body}
+			config, format, err := image.DecodeConfig(reader)
+			assert.NoError(t, err)
+			assert.Equal(t, c.format, format)
+			assert.Equal(t, c.width, config.Width)
+			assert.Equal(t, c.height, config.Height)
+			t.Logf("Bytes read: %d", reader.BytesRead)
+		})
+	}
+}
+
+func TestBase64(t *testing.T) {
+	// Bytes read:
+	// jpeg: 1063892
+	// png: 294462
+	// webp: 99072
+	// gif: 953856
+	// jpeg#01: 32805
+	for _, c := range cases {
+		t.Run("Decode:"+c.format, func(t *testing.T) {
+			resp, err := http.Get(c.url)
+			assert.NoError(t, err)
+			defer resp.Body.Close()
+			data, err := io.ReadAll(resp.Body)
+			assert.NoError(t, err)
+			encoded := base64.StdEncoding.EncodeToString(data)
+			body := base64.NewDecoder(base64.StdEncoding, strings.NewReader(encoded))
+			reader := &CountingReader{reader: body}
+			img, format, err := image.Decode(reader)
+			assert.NoError(t, err)
+			size := img.Bounds().Size()
+			assert.Equal(t, c.format, format)
+			assert.Equal(t, c.width, size.X)
+			assert.Equal(t, c.height, size.Y)
+			t.Logf("Bytes read: %d", reader.BytesRead)
+		})
+	}
+
+	// Bytes read:
+	// jpeg: 1536
+	// png: 768
+	// webp: 768
+	// gif: 1536
+	// jpeg#01: 3840
+	for _, c := range cases {
+		t.Run("DecodeConfig:"+c.format, func(t *testing.T) {
+			resp, err := http.Get(c.url)
+			assert.NoError(t, err)
+			defer resp.Body.Close()
+			data, err := io.ReadAll(resp.Body)
+			assert.NoError(t, err)
+			encoded := base64.StdEncoding.EncodeToString(data)
+			body := base64.NewDecoder(base64.StdEncoding, strings.NewReader(encoded))
+			reader := &CountingReader{reader: body}
+			config, format, err := image.DecodeConfig(reader)
+			assert.NoError(t, err)
+			assert.Equal(t, c.format, format)
+			assert.Equal(t, c.width, config.Width)
+			assert.Equal(t, c.height, config.Height)
+			t.Logf("Bytes read: %d", reader.BytesRead)
+		})
+	}
+}
+
+func TestGetImageSize(t *testing.T) {
+	for i, c := range cases {
+		t.Run("Decode:"+strconv.Itoa(i), func(t *testing.T) {
+			width, height, err := img.GetImageSize(c.url)
+			assert.NoError(t, err)
+			assert.Equal(t, c.width, width)
+			assert.Equal(t, c.height, height)
+		})
+	}
+}

common/init.go

@@ -12,7 +12,7 @@ var (
 	Port         = flag.Int("port", 3000, "the listening port")
 	PrintVersion = flag.Bool("version", false, "print version and exit")
 	PrintHelp    = flag.Bool("help", false, "print help and exit")
-	LogDir       = flag.String("log-dir", "", "specify the log directory")
+	LogDir       = flag.String("log-dir", "./logs", "specify the log directory")
 )
 
 func printHelp() {

common/logger.go

@@ -1,29 +1,47 @@
 package common
 
 import (
+	"context"
 	"fmt"
 	"github.com/gin-gonic/gin"
 	"io"
 	"log"
 	"os"
 	"path/filepath"
+	"sync"
 	"time"
 )
 
-func SetupGinLog() {
+const (
+	loggerINFO  = "INFO"
+	loggerWarn  = "WARN"
+	loggerError = "ERR"
+)
+
+const maxLogCount = 1000000
+
+var logCount int
+var setupLogLock sync.Mutex
+var setupLogWorking bool
+
+func SetupLogger() {
 	if *LogDir != "" {
-		commonLogPath := filepath.Join(*LogDir, "common.log")
-		errorLogPath := filepath.Join(*LogDir, "error.log")
-		commonFd, err := os.OpenFile(commonLogPath, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
+		ok := setupLogLock.TryLock()
+		if !ok {
+			log.Println("setup log is already working")
+			return
+		}
+		defer func() {
+			setupLogLock.Unlock()
+			setupLogWorking = false
+		}()
+		logPath := filepath.Join(*LogDir, fmt.Sprintf("oneapi-%s.log", time.Now().Format("20060102")))
+		fd, err := os.OpenFile(logPath, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
 		if err != nil {
 			log.Fatal("failed to open log file")
 		}
-		errorFd, err := os.OpenFile(errorLogPath, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644)
-		if err != nil {
-			log.Fatal("failed to open log file")
-		}
-		gin.DefaultWriter = io.MultiWriter(os.Stdout, commonFd)
-		gin.DefaultErrorWriter = io.MultiWriter(os.Stderr, errorFd)
+		gin.DefaultWriter = io.MultiWriter(os.Stdout, fd)
+		gin.DefaultErrorWriter = io.MultiWriter(os.Stderr, fd)
 	}
 }
 
@@ -37,6 +55,36 @@ func SysError(s string) {
 	_, _ = fmt.Fprintf(gin.DefaultErrorWriter, "[SYS] %v | %s \n", t.Format("2006/01/02 - 15:04:05"), s)
 }
 
+func LogInfo(ctx context.Context, msg string) {
+	logHelper(ctx, loggerINFO, msg)
+}
+
+func LogWarn(ctx context.Context, msg string) {
+	logHelper(ctx, loggerWarn, msg)
+}
+
+func LogError(ctx context.Context, msg string) {
+	logHelper(ctx, loggerError, msg)
+}
+
+func logHelper(ctx context.Context, level string, msg string) {
+	writer := gin.DefaultErrorWriter
+	if level == loggerINFO {
+		writer = gin.DefaultWriter
+	}
+	id := ctx.Value(RequestIdKey)
+	now := time.Now()
+	_, _ = fmt.Fprintf(writer, "[%s] %v | %s | %s \n", level, now.Format("2006/01/02 - 15:04:05"), id, msg)
+	logCount++ // we don't need accurate count, so no lock here
+	if logCount > maxLogCount && !setupLogWorking {
+		logCount = 0
+		setupLogWorking = true
+		go func() {
+			SetupLogger()
+		}()
+	}
+}
+
 func FatalLog(v ...any) {
 	t := time.Now()
 	_, _ = fmt.Fprintf(gin.DefaultErrorWriter, "[FATAL] %v | %v \n", t.Format("2006/01/02 - 15:04:05"), v)

common/model-ratio.go

@@ -3,8 +3,32 @@ package common
 import (
 	"encoding/json"
 	"strings"
+	"time"
 )
 
+var DalleSizeRatios = map[string]map[string]float64{
+	"dall-e-2": {
+		"256x256":   1,
+		"512x512":   1.125,
+		"1024x1024": 1.25,
+	},
+	"dall-e-3": {
+		"1024x1024": 1,
+		"1024x1792": 2,
+		"1792x1024": 2,
+	},
+}
+
+var DalleGenerationImageAmounts = map[string][2]int{
+	"dall-e-2": {1, 10},
+	"dall-e-3": {1, 1}, // OpenAI allows n=1 currently.
+}
+
+var DalleImagePromptLengthLimitations = map[string]int{
+	"dall-e-2": 1000,
+	"dall-e-3": 4000,
+}
+
 // ModelRatio
 // https://platform.openai.com/docs/models/model-endpoint-compatibility
 // https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Blfmc9dlf
@@ -19,11 +43,15 @@ var ModelRatio = map[string]float64{
 	"gpt-4-32k":                 30,
 	"gpt-4-32k-0314":            30,
 	"gpt-4-32k-0613":            30,
+	"gpt-4-1106-preview":        5,    // $0.01 / 1K tokens
+	"gpt-4-vision-preview":      5,    // $0.01 / 1K tokens
 	"gpt-3.5-turbo":             0.75, // $0.0015 / 1K tokens
 	"gpt-3.5-turbo-0301":        0.75,
 	"gpt-3.5-turbo-0613":        0.75,
 	"gpt-3.5-turbo-16k":         1.5, // $0.003 / 1K tokens
 	"gpt-3.5-turbo-16k-0613":    1.5,
+	"gpt-3.5-turbo-instruct":    0.75, // $0.0015 / 1K tokens
+	"gpt-3.5-turbo-1106":        0.5,  // $0.001 / 1K tokens
 	"text-ada-001":              0.2,
 	"text-babbage-001":          0.25,
 	"text-curie-001":            1,
@@ -31,7 +59,11 @@ var ModelRatio = map[string]float64{
 	"text-davinci-003":          10,
 	"text-davinci-edit-001":     10,
 	"code-davinci-edit-001":     10,
-	"whisper-1":                 10,
+	"whisper-1":                 15,  // $0.006 / minute -> $0.006 / 150 words -> $0.006 / 200 tokens -> $0.03 / 1k tokens
+	"tts-1":                     7.5, // $0.015 / 1K characters
+	"tts-1-1106":                7.5,
+	"tts-1-hd":                  15, // $0.030 / 1K characters
+	"tts-1-hd-1106":             15,
 	"davinci":                   10,
 	"curie":                     10,
 	"babbage":                   10,
@@ -40,24 +72,31 @@ var ModelRatio = map[string]float64{
 	"text-search-ada-doc-001":   10,
 	"text-moderation-stable":    0.1,
 	"text-moderation-latest":    0.1,
-	"dall-e":                    8,
+	"dall-e-2":                  8,      // $0.016 - $0.020 / image
+	"dall-e-3":                  20,     // $0.040 - $0.120 / image
 	"claude-instant-1":          0.815,  // $1.63 / 1M tokens
 	"claude-2":                  5.51,   // $11.02 / 1M tokens
+	"claude-2.0":                5.51,   // $11.02 / 1M tokens
+	"claude-2.1":                5.51,   // $11.02 / 1M tokens
 	"ERNIE-Bot":                 0.8572, // ￥0.012 / 1k tokens
 	"ERNIE-Bot-turbo":           0.5715, // ￥0.008 / 1k tokens
+	"ERNIE-Bot-4":               8.572,  // ￥0.12 / 1k tokens
 	"Embedding-V1":              0.1429, // ￥0.002 / 1k tokens
 	"PaLM-2":                    1,
+	"gemini-pro":                1,      // $0.00025 / 1k characters -> $0.001 / 1k tokens
+	"chatglm_turbo":             0.3572, // ￥0.005 / 1k tokens
 	"chatglm_pro":               0.7143, // ￥0.01 / 1k tokens
 	"chatglm_std":               0.3572, // ￥0.005 / 1k tokens
 	"chatglm_lite":              0.1429, // ￥0.002 / 1k tokens
-	"qwen-v1":                   0.8572, // TBD: https://help.aliyun.com/document_detail/2399482.html?spm=a2c4g.2399482.0.0.1ad347feilAgag
-	"qwen-plus-v1":              0.5715, // Same as above
-	"SparkDesk":                 0.8572, // TBD
+	"qwen-turbo":                0.8572, // ￥0.012 / 1k tokens
+	"qwen-plus":                 10,     // ￥0.14 / 1k tokens
+	"text-embedding-v1":         0.05,   // ￥0.0007 / 1k tokens
+	"SparkDesk":                 1.2858, // ￥0.018 / 1k tokens
 	"360GPT_S2_V9":              0.8572, // ¥0.012 / 1k tokens
 	"embedding-bert-512-v1":     0.0715, // ¥0.001 / 1k tokens
 	"embedding_s1_v1":           0.0715, // ¥0.001 / 1k tokens
 	"semantic_similarity_s1_v1": 0.0715, // ¥0.001 / 1k tokens
-	"360GPT_S2_V9.4":            0.8572, // ¥0.012 / 1k tokens
+	"hunyuan":                   7.143,  // ¥0.1 / 1k tokens  // https://cloud.tencent.com/document/product/1729/97731#e0e6be58-60c8-469f-bdeb-6c264ce3b4d0
 }
 
 func ModelRatio2JSONString() string {
@@ -84,9 +123,24 @@ func GetModelRatio(name string) float64 {
 
 func GetCompletionRatio(name string) float64 {
 	if strings.HasPrefix(name, "gpt-3.5") {
+		if strings.HasSuffix(name, "1106") {
+			return 2
+		}
+		if name == "gpt-3.5-turbo" || name == "gpt-3.5-turbo-16k" {
+			// TODO: clear this after 2023-12-11
+			now := time.Now()
+			// https://platform.openai.com/docs/models/continuous-model-upgrades
+			// if after 2023-12-11, use 2
+			if now.After(time.Date(2023, 12, 11, 0, 0, 0, 0, time.UTC)) {
+				return 2
+			}
+		}
 		return 1.333333
 	}
 	if strings.HasPrefix(name, "gpt-4") {
+		if strings.HasSuffix(name, "preview") {
+			return 3
+		}
 		return 2
 	}
 	if strings.HasPrefix(name, "claude-instant-1") {

common/utils.go

@@ -171,6 +171,11 @@ func GetTimestamp() int64 {
 	return time.Now().Unix()
 }
 
+func GetTimeString() string {
+	now := time.Now()
+	return fmt.Sprintf("%s%d", now.Format("20060102150405"), now.UnixNano()%1e9)
+}
+
 func Max(a int, b int) int {
 	if a >= b {
 		return a
@@ -190,3 +195,15 @@ func GetOrDefault(env string, defaultValue int) int {
 	}
 	return num
 }
+
+func MessageWithRequestId(message string, id string) string {
+	return fmt.Sprintf("%s (request id: %s)", message, id)
+}
+
+func String2Int(str string) int {
+	num, err := strconv.Atoi(str)
+	if err != nil {
+		return 0
+	}
+	return num
+}

controller/billing.go

@@ -29,7 +29,7 @@ func GetSubscription(c *gin.Context) {
 	if err != nil {
 		openAIError := OpenAIError{
 			Message: err.Error(),
-			Type:    "one_api_error",
+			Type:    "upstream_error",
 		}
 		c.JSON(200, gin.H{
 			"error": openAIError,

controller/channel-billing.go

@@ -111,7 +111,7 @@ func GetResponseBody(method, url string, channel *model.Channel, headers http.He
 }
 
 func updateChannelCloseAIBalance(channel *model.Channel) (float64, error) {
-	url := fmt.Sprintf("%s/dashboard/billing/credit_grants", channel.BaseURL)
+	url := fmt.Sprintf("%s/dashboard/billing/credit_grants", channel.GetBaseURL())
 	body, err := GetResponseBody("GET", url, channel, GetAuthHeader(channel.Key))
 
 	if err != nil {
@@ -201,18 +201,18 @@ func updateChannelAIGC2DBalance(channel *model.Channel) (float64, error) {
 
 func updateChannelBalance(channel *model.Channel) (float64, error) {
 	baseURL := common.ChannelBaseURLs[channel.Type]
-	if channel.BaseURL == "" {
-		channel.BaseURL = baseURL
+	if channel.GetBaseURL() == "" {
+		channel.BaseURL = &baseURL
 	}
 	switch channel.Type {
 	case common.ChannelTypeOpenAI:
-		if channel.BaseURL != "" {
-			baseURL = channel.BaseURL
+		if channel.GetBaseURL() != "" {
+			baseURL = channel.GetBaseURL()
 		}
 	case common.ChannelTypeAzure:
 		return 0, errors.New("尚未实现")
 	case common.ChannelTypeCustom:
-		baseURL = channel.BaseURL
+		baseURL = channel.GetBaseURL()
 	case common.ChannelTypeCloseAI:
 		return updateChannelCloseAIBalance(channel)
 	case common.ChannelTypeOpenAISB:

controller/channel-test.go

@@ -5,19 +5,23 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
+	"io"
 	"net/http"
 	"one-api/common"
 	"one-api/model"
 	"strconv"
 	"sync"
 	"time"
+
+	"github.com/gin-gonic/gin"
 )
 
-func testChannel(channel *model.Channel, request ChatRequest) (error, *OpenAIError) {
+func testChannel(channel *model.Channel, request ChatRequest) (err error, openaiErr *OpenAIError) {
 	switch channel.Type {
 	case common.ChannelTypePaLM:
 		fallthrough
+	case common.ChannelTypeGemini:
+		fallthrough
 	case common.ChannelTypeAnthropic:
 		fallthrough
 	case common.ChannelTypeBaidu:
@@ -32,19 +36,24 @@ func testChannel(channel *model.Channel, request ChatRequest) (error, *OpenAIErr
 		return errors.New("该渠道类型当前版本不支持测试，请手动测试"), nil
 	case common.ChannelTypeAzure:
 		request.Model = "gpt-35-turbo"
+		defer func() {
+			if err != nil {
+				err = errors.New("请确保已在 Azure 上创建了 gpt-35-turbo 模型，并且 apiVersion 已正确填写！")
+			}
+		}()
 	default:
 		request.Model = "gpt-3.5-turbo"
 	}
 	requestURL := common.ChannelBaseURLs[channel.Type]
 	if channel.Type == common.ChannelTypeAzure {
-		requestURL = fmt.Sprintf("%s/openai/deployments/%s/chat/completions?api-version=2023-03-15-preview", channel.BaseURL, request.Model)
+		requestURL = getFullRequestURL(channel.GetBaseURL(), fmt.Sprintf("/openai/deployments/%s/chat/completions?api-version=2023-03-15-preview", request.Model), channel.Type)
 	} else {
-		if channel.BaseURL != "" {
-			requestURL = channel.BaseURL
+		if baseURL := channel.GetBaseURL(); len(baseURL) > 0 {
+			requestURL = baseURL
 		}
-		requestURL += "/v1/chat/completions"
-	}
 
+		requestURL = getFullRequestURL(requestURL, "/v1/chat/completions", channel.Type)
+	}
 	jsonData, err := json.Marshal(request)
 	if err != nil {
 		return err, nil
@@ -65,11 +74,18 @@ func testChannel(channel *model.Channel, request ChatRequest) (error, *OpenAIErr
 	}
 	defer resp.Body.Close()
 	var response TextResponse
-	err = json.NewDecoder(resp.Body).Decode(&response)
+	body, err := io.ReadAll(resp.Body)
 	if err != nil {
 		return err, nil
 	}
+	err = json.Unmarshal(body, &response)
+	if err != nil {
+		return fmt.Errorf("Error: %s\nResp body: %s", err, body), nil
+	}
 	if response.Usage.CompletionTokens == 0 {
+		if response.Error.Message == "" {
+			response.Error.Message = "补全 tokens 非预期返回 0"
+		}
 		return errors.New(fmt.Sprintf("type %s, code %v, message %s", response.Error.Type, response.Error.Code, response.Error.Message)), &response.Error
 	}
 	return nil, nil
@@ -131,20 +147,32 @@ func TestChannel(c *gin.Context) {
 var testAllChannelsLock sync.Mutex
 var testAllChannelsRunning bool = false
 
-// disable & notify
-func disableChannel(channelId int, channelName string, reason string) {
+func notifyRootUser(subject string, content string) {
 	if common.RootUserEmail == "" {
 		common.RootUserEmail = model.GetRootUserEmail()
 	}
-	model.UpdateChannelStatusById(channelId, common.ChannelStatusDisabled)
-	subject := fmt.Sprintf("通道「%s」（#%d）已被禁用", channelName, channelId)
-	content := fmt.Sprintf("通道「%s」（#%d）已被禁用，原因：%s", channelName, channelId, reason)
 	err := common.SendEmail(subject, common.RootUserEmail, content)
 	if err != nil {
 		common.SysError(fmt.Sprintf("failed to send email: %s", err.Error()))
 	}
 }
 
+// disable & notify
+func disableChannel(channelId int, channelName string, reason string) {
+	model.UpdateChannelStatusById(channelId, common.ChannelStatusAutoDisabled)
+	subject := fmt.Sprintf("通道「%s」（#%d）已被禁用", channelName, channelId)
+	content := fmt.Sprintf("通道「%s」（#%d）已被禁用，原因：%s", channelName, channelId, reason)
+	notifyRootUser(subject, content)
+}
+
+// enable & notify
+func enableChannel(channelId int, channelName string) {
+	model.UpdateChannelStatusById(channelId, common.ChannelStatusEnabled)
+	subject := fmt.Sprintf("通道「%s」（#%d）已被启用", channelName, channelId)
+	content := fmt.Sprintf("通道「%s」（#%d）已被启用", channelName, channelId)
+	notifyRootUser(subject, content)
+}
+
 func testAllChannels(notify bool) error {
 	if common.RootUserEmail == "" {
 		common.RootUserEmail = model.GetRootUserEmail()
@@ -167,20 +195,21 @@ func testAllChannels(notify bool) error {
 	}
 	go func() {
 		for _, channel := range channels {
-			if channel.Status != common.ChannelStatusEnabled {
-				continue
-			}
+			isChannelEnabled := channel.Status == common.ChannelStatusEnabled
 			tik := time.Now()
 			err, openaiErr := testChannel(channel, *testRequest)
 			tok := time.Now()
 			milliseconds := tok.Sub(tik).Milliseconds()
-			if milliseconds > disableThreshold {
+			if isChannelEnabled && milliseconds > disableThreshold {
 				err = errors.New(fmt.Sprintf("响应时间 %.2fs 超过阈值 %.2fs", float64(milliseconds)/1000.0, float64(disableThreshold)/1000.0))
 				disableChannel(channel.Id, channel.Name, err.Error())
 			}
-			if shouldDisableChannel(openaiErr, -1) {
+			if isChannelEnabled && shouldDisableChannel(openaiErr, -1) {
 				disableChannel(channel.Id, channel.Name, err.Error())
 			}
+			if !isChannelEnabled && shouldEnableChannel(err, openaiErr) {
+				enableChannel(channel.Id, channel.Name)
+			}
 			channel.UpdateResponseTime(milliseconds)
 			time.Sleep(common.RequestInterval)
 		}

controller/channel.go

@@ -85,7 +85,7 @@ func AddChannel(c *gin.Context) {
 	}
 	channel.CreatedTime = common.GetTimestamp()
 	keys := strings.Split(channel.Key, "\n")
-	channels := make([]model.Channel, 0)
+	channels := make([]model.Channel, 0, len(keys))
 	for _, key := range keys {
 		if key == "" {
 			continue
@@ -127,6 +127,23 @@ func DeleteChannel(c *gin.Context) {
 	return
 }
 
+func DeleteDisabledChannel(c *gin.Context) {
+	rows, err := model.DeleteDisabledChannel()
+	if err != nil {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": err.Error(),
+		})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{
+		"success": true,
+		"message": "",
+		"data":    rows,
+	})
+	return
+}
+
 func UpdateChannel(c *gin.Context) {
 	channel := model.Channel{}
 	err := c.ShouldBindJSON(&channel)

controller/github.go

@@ -79,6 +79,14 @@ func getGitHubUserInfoByCode(code string) (*GitHubUser, error) {
 
 func GitHubOAuth(c *gin.Context) {
 	session := sessions.Default(c)
+	state := c.Query("state")
+	if state == "" || session.Get("oauth_state") == nil || state != session.Get("oauth_state").(string) {
+		c.JSON(http.StatusForbidden, gin.H{
+			"success": false,
+			"message": "state is empty or not same",
+		})
+		return
+	}
 	username := session.Get("username")
 	if username != nil {
 		GitHubBind(c)
@@ -205,3 +213,22 @@ func GitHubBind(c *gin.Context) {
 	})
 	return
 }
+
+func GenerateOAuthCode(c *gin.Context) {
+	session := sessions.Default(c)
+	state := common.GetRandomString(12)
+	session.Set("oauth_state", state)
+	err := session.Save()
+	if err != nil {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": err.Error(),
+		})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{
+		"success": true,
+		"message": "",
+		"data":    state,
+	})
+}

controller/log.go

@@ -2,6 +2,7 @@ package controller
 
 import (
 	"github.com/gin-gonic/gin"
+	"net/http"
 	"one-api/common"
 	"one-api/model"
 	"strconv"
@@ -18,19 +19,21 @@ func GetAllLogs(c *gin.Context) {
 	username := c.Query("username")
 	tokenName := c.Query("token_name")
 	modelName := c.Query("model_name")
-	logs, err := model.GetAllLogs(logType, startTimestamp, endTimestamp, modelName, username, tokenName, p*common.ItemsPerPage, common.ItemsPerPage)
+	channel, _ := strconv.Atoi(c.Query("channel"))
+	logs, err := model.GetAllLogs(logType, startTimestamp, endTimestamp, modelName, username, tokenName, p*common.ItemsPerPage, common.ItemsPerPage, channel)
 	if err != nil {
-		c.JSON(200, gin.H{
+		c.JSON(http.StatusOK, gin.H{
 			"success": false,
 			"message": err.Error(),
 		})
 		return
 	}
-	c.JSON(200, gin.H{
+	c.JSON(http.StatusOK, gin.H{
 		"success": true,
 		"message": "",
 		"data":    logs,
 	})
+	return
 }
 
 func GetUserLogs(c *gin.Context) {
@@ -46,34 +49,36 @@ func GetUserLogs(c *gin.Context) {
 	modelName := c.Query("model_name")
 	logs, err := model.GetUserLogs(userId, logType, startTimestamp, endTimestamp, modelName, tokenName, p*common.ItemsPerPage, common.ItemsPerPage)
 	if err != nil {
-		c.JSON(200, gin.H{
+		c.JSON(http.StatusOK, gin.H{
 			"success": false,
 			"message": err.Error(),
 		})
 		return
 	}
-	c.JSON(200, gin.H{
+	c.JSON(http.StatusOK, gin.H{
 		"success": true,
 		"message": "",
 		"data":    logs,
 	})
+	return
 }
 
 func SearchAllLogs(c *gin.Context) {
 	keyword := c.Query("keyword")
 	logs, err := model.SearchAllLogs(keyword)
 	if err != nil {
-		c.JSON(200, gin.H{
+		c.JSON(http.StatusOK, gin.H{
 			"success": false,
 			"message": err.Error(),
 		})
 		return
 	}
-	c.JSON(200, gin.H{
+	c.JSON(http.StatusOK, gin.H{
 		"success": true,
 		"message": "",
 		"data":    logs,
 	})
+	return
 }
 
 func SearchUserLogs(c *gin.Context) {
@@ -81,17 +86,18 @@ func SearchUserLogs(c *gin.Context) {
 	userId := c.GetInt("id")
 	logs, err := model.SearchUserLogs(userId, keyword)
 	if err != nil {
-		c.JSON(200, gin.H{
+		c.JSON(http.StatusOK, gin.H{
 			"success": false,
 			"message": err.Error(),
 		})
 		return
 	}
-	c.JSON(200, gin.H{
+	c.JSON(http.StatusOK, gin.H{
 		"success": true,
 		"message": "",
 		"data":    logs,
 	})
+	return
 }
 
 func GetLogsStat(c *gin.Context) {
@@ -101,9 +107,10 @@ func GetLogsStat(c *gin.Context) {
 	tokenName := c.Query("token_name")
 	username := c.Query("username")
 	modelName := c.Query("model_name")
-	quotaNum := model.SumUsedQuota(logType, startTimestamp, endTimestamp, modelName, username, tokenName)
+	channel, _ := strconv.Atoi(c.Query("channel"))
+	quotaNum := model.SumUsedQuota(logType, startTimestamp, endTimestamp, modelName, username, tokenName, channel)
 	//tokenNum := model.SumUsedToken(logType, startTimestamp, endTimestamp, modelName, username, "")
-	c.JSON(200, gin.H{
+	c.JSON(http.StatusOK, gin.H{
 		"success": true,
 		"message": "",
 		"data": gin.H{
@@ -111,6 +118,7 @@ func GetLogsStat(c *gin.Context) {
 			//"token": tokenNum,
 		},
 	})
+	return
 }
 
 func GetLogsSelfStat(c *gin.Context) {
@@ -120,9 +128,10 @@ func GetLogsSelfStat(c *gin.Context) {
 	endTimestamp, _ := strconv.ParseInt(c.Query("end_timestamp"), 10, 64)
 	tokenName := c.Query("token_name")
 	modelName := c.Query("model_name")
-	quotaNum := model.SumUsedQuota(logType, startTimestamp, endTimestamp, modelName, username, tokenName)
+	channel, _ := strconv.Atoi(c.Query("channel"))
+	quotaNum := model.SumUsedQuota(logType, startTimestamp, endTimestamp, modelName, username, tokenName, channel)
 	//tokenNum := model.SumUsedToken(logType, startTimestamp, endTimestamp, modelName, username, tokenName)
-	c.JSON(200, gin.H{
+	c.JSON(http.StatusOK, gin.H{
 		"success": true,
 		"message": "",
 		"data": gin.H{
@@ -130,4 +139,30 @@ func GetLogsSelfStat(c *gin.Context) {
 			//"token": tokenNum,
 		},
 	})
+	return
+}
+
+func DeleteHistoryLogs(c *gin.Context) {
+	targetTimestamp, _ := strconv.ParseInt(c.Query("target_timestamp"), 10, 64)
+	if targetTimestamp == 0 {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": "target timestamp is required",
+		})
+		return
+	}
+	count, err := model.DeleteOldLog(targetTimestamp)
+	if err != nil {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": err.Error(),
+		})
+		return
+	}
+	c.JSON(http.StatusOK, gin.H{
+		"success": true,
+		"message": "",
+		"data":    count,
+	})
+	return
 }

controller/model.go

@@ -55,12 +55,66 @@ func init() {
 	// https://platform.openai.com/docs/models/model-endpoint-compatibility
 	openAIModels = []OpenAIModels{
 		{
-			Id:         "dall-e",
+			Id:         "dall-e-2",
 			Object:     "model",
 			Created:    1677649963,
 			OwnedBy:    "openai",
 			Permission: permission,
-			Root:       "dall-e",
+			Root:       "dall-e-2",
+			Parent:     nil,
+		},
+		{
+			Id:         "dall-e-3",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "dall-e-3",
+			Parent:     nil,
+		},
+		{
+			Id:         "whisper-1",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "whisper-1",
+			Parent:     nil,
+		},
+		{
+			Id:         "tts-1",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "tts-1",
+			Parent:     nil,
+		},
+		{
+			Id:         "tts-1-1106",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "tts-1-1106",
+			Parent:     nil,
+		},
+		{
+			Id:         "tts-1-hd",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "tts-1-hd",
+			Parent:     nil,
+		},
+		{
+			Id:         "tts-1-hd-1106",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "tts-1-hd-1106",
 			Parent:     nil,
 		},
 		{
@@ -108,6 +162,24 @@ func init() {
 			Root:       "gpt-3.5-turbo-16k-0613",
 			Parent:     nil,
 		},
+		{
+			Id:         "gpt-3.5-turbo-1106",
+			Object:     "model",
+			Created:    1699593571,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "gpt-3.5-turbo-1106",
+			Parent:     nil,
+		},
+		{
+			Id:         "gpt-3.5-turbo-instruct",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "gpt-3.5-turbo-instruct",
+			Parent:     nil,
+		},
 		{
 			Id:         "gpt-4",
 			Object:     "model",
@@ -162,6 +234,24 @@ func init() {
 			Root:       "gpt-4-32k-0613",
 			Parent:     nil,
 		},
+		{
+			Id:         "gpt-4-1106-preview",
+			Object:     "model",
+			Created:    1699593571,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "gpt-4-1106-preview",
+			Parent:     nil,
+		},
+		{
+			Id:         "gpt-4-vision-preview",
+			Object:     "model",
+			Created:    1699593571,
+			OwnedBy:    "openai",
+			Permission: permission,
+			Root:       "gpt-4-vision-preview",
+			Parent:     nil,
+		},
 		{
 			Id:         "text-embedding-ada-002",
 			Object:     "model",
@@ -256,7 +346,7 @@ func init() {
 			Id:         "claude-instant-1",
 			Object:     "model",
 			Created:    1677649963,
-			OwnedBy:    "anturopic",
+			OwnedBy:    "anthropic",
 			Permission: permission,
 			Root:       "claude-instant-1",
 			Parent:     nil,
@@ -265,11 +355,29 @@ func init() {
 			Id:         "claude-2",
 			Object:     "model",
 			Created:    1677649963,
-			OwnedBy:    "anturopic",
+			OwnedBy:    "anthropic",
 			Permission: permission,
 			Root:       "claude-2",
 			Parent:     nil,
 		},
+		{
+			Id:         "claude-2.1",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "anthropic",
+			Permission: permission,
+			Root:       "claude-2.1",
+			Parent:     nil,
+		},
+		{
+			Id:         "claude-2.0",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "anthropic",
+			Permission: permission,
+			Root:       "claude-2.0",
+			Parent:     nil,
+		},
 		{
 			Id:         "ERNIE-Bot",
 			Object:     "model",
@@ -288,6 +396,15 @@ func init() {
 			Root:       "ERNIE-Bot-turbo",
 			Parent:     nil,
 		},
+		{
+			Id:         "ERNIE-Bot-4",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "baidu",
+			Permission: permission,
+			Root:       "ERNIE-Bot-4",
+			Parent:     nil,
+		},
 		{
 			Id:         "Embedding-V1",
 			Object:     "model",
@@ -306,6 +423,24 @@ func init() {
 			Root:       "PaLM-2",
 			Parent:     nil,
 		},
+		{
+			Id:         "gemini-pro",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "google",
+			Permission: permission,
+			Root:       "gemini-pro",
+			Parent:     nil,
+		},
+		{
+			Id:         "chatglm_turbo",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "zhipu",
+			Permission: permission,
+			Root:       "chatglm_turbo",
+			Parent:     nil,
+		},
 		{
 			Id:         "chatglm_pro",
 			Object:     "model",
@@ -334,21 +469,30 @@ func init() {
 			Parent:     nil,
 		},
 		{
-			Id:         "qwen-v1",
+			Id:         "qwen-turbo",
 			Object:     "model",
 			Created:    1677649963,
 			OwnedBy:    "ali",
 			Permission: permission,
-			Root:       "qwen-v1",
+			Root:       "qwen-turbo",
 			Parent:     nil,
 		},
 		{
-			Id:         "qwen-plus-v1",
+			Id:         "qwen-plus",
 			Object:     "model",
 			Created:    1677649963,
 			OwnedBy:    "ali",
 			Permission: permission,
-			Root:       "qwen-plus-v1",
+			Root:       "qwen-plus",
+			Parent:     nil,
+		},
+		{
+			Id:         "text-embedding-v1",
+			Object:     "model",
+			Created:    1677649963,
+			OwnedBy:    "ali",
+			Permission: permission,
+			Root:       "text-embedding-v1",
 			Parent:     nil,
 		},
 		{
@@ -397,12 +541,12 @@ func init() {
 			Parent:     nil,
 		},
 		{
-			Id:         "360GPT_S2_V9.4",
+			Id:         "hunyuan",
 			Object:     "model",
 			Created:    1677649963,
-			OwnedBy:    "360",
+			OwnedBy:    "tencent",
 			Permission: permission,
-			Root:       "360GPT_S2_V9.4",
+			Root:       "hunyuan",
 			Parent:     nil,
 		},
 	}

controller/option.go

@@ -46,7 +46,7 @@ func UpdateOption(c *gin.Context) {
 		if option.Value == "true" && common.GitHubClientId == "" {
 			c.JSON(http.StatusOK, gin.H{
 				"success": false,
-				"message": "无法启用 GitHub OAuth，请先填入 GitHub Client ID 以及 GitHub Client Secret！",
+				"message": "无法启用 GitHub OAuth，请先填入 GitHub Client Id 以及 GitHub Client Secret！",
 			})
 			return
 		}

controller/relay-aiproxy.go

@@ -0,0 +1,220 @@
+package controller
+
+import (
+	"bufio"
+	"encoding/json"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"strconv"
+	"strings"
+)
+
+// https://docs.aiproxy.io/dev/library#使用已经定制好的知识库进行对话问答
+
+type AIProxyLibraryRequest struct {
+	Model     string `json:"model"`
+	Query     string `json:"query"`
+	LibraryId string `json:"libraryId"`
+	Stream    bool   `json:"stream"`
+}
+
+type AIProxyLibraryError struct {
+	ErrCode int    `json:"errCode"`
+	Message string `json:"message"`
+}
+
+type AIProxyLibraryDocument struct {
+	Title string `json:"title"`
+	URL   string `json:"url"`
+}
+
+type AIProxyLibraryResponse struct {
+	Success   bool                     `json:"success"`
+	Answer    string                   `json:"answer"`
+	Documents []AIProxyLibraryDocument `json:"documents"`
+	AIProxyLibraryError
+}
+
+type AIProxyLibraryStreamResponse struct {
+	Content   string                   `json:"content"`
+	Finish    bool                     `json:"finish"`
+	Model     string                   `json:"model"`
+	Documents []AIProxyLibraryDocument `json:"documents"`
+}
+
+func requestOpenAI2AIProxyLibrary(request GeneralOpenAIRequest) *AIProxyLibraryRequest {
+	query := ""
+	if len(request.Messages) != 0 {
+		query = request.Messages[len(request.Messages)-1].StringContent()
+	}
+	return &AIProxyLibraryRequest{
+		Model:  request.Model,
+		Stream: request.Stream,
+		Query:  query,
+	}
+}
+
+func aiProxyDocuments2Markdown(documents []AIProxyLibraryDocument) string {
+	if len(documents) == 0 {
+		return ""
+	}
+	content := "\n\n参考文档：\n"
+	for i, document := range documents {
+		content += fmt.Sprintf("%d. [%s](%s)\n", i+1, document.Title, document.URL)
+	}
+	return content
+}
+
+func responseAIProxyLibrary2OpenAI(response *AIProxyLibraryResponse) *OpenAITextResponse {
+	content := response.Answer + aiProxyDocuments2Markdown(response.Documents)
+	choice := OpenAITextResponseChoice{
+		Index: 0,
+		Message: Message{
+			Role:    "assistant",
+			Content: content,
+		},
+		FinishReason: "stop",
+	}
+	fullTextResponse := OpenAITextResponse{
+		Id:      common.GetUUID(),
+		Object:  "chat.completion",
+		Created: common.GetTimestamp(),
+		Choices: []OpenAITextResponseChoice{choice},
+	}
+	return &fullTextResponse
+}
+
+func documentsAIProxyLibrary(documents []AIProxyLibraryDocument) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = aiProxyDocuments2Markdown(documents)
+	choice.FinishReason = &stopFinishReason
+	return &ChatCompletionsStreamResponse{
+		Id:      common.GetUUID(),
+		Object:  "chat.completion.chunk",
+		Created: common.GetTimestamp(),
+		Model:   "",
+		Choices: []ChatCompletionsStreamResponseChoice{choice},
+	}
+}
+
+func streamResponseAIProxyLibrary2OpenAI(response *AIProxyLibraryStreamResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	choice.Delta.Content = response.Content
+	return &ChatCompletionsStreamResponse{
+		Id:      common.GetUUID(),
+		Object:  "chat.completion.chunk",
+		Created: common.GetTimestamp(),
+		Model:   response.Model,
+		Choices: []ChatCompletionsStreamResponseChoice{choice},
+	}
+}
+
+func aiProxyLibraryStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var usage Usage
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
+		if atEOF && len(data) == 0 {
+			return 0, nil, nil
+		}
+		if i := strings.Index(string(data), "\n"); i >= 0 {
+			return i + 1, data[0:i], nil
+		}
+		if atEOF {
+			return len(data), data, nil
+		}
+		return 0, nil, nil
+	})
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		for scanner.Scan() {
+			data := scanner.Text()
+			if len(data) < 5 { // ignore blank line or wrong format
+				continue
+			}
+			if data[:5] != "data:" {
+				continue
+			}
+			data = data[5:]
+			dataChan <- data
+		}
+		stopChan <- true
+	}()
+	setEventStreamHeaders(c)
+	var documents []AIProxyLibraryDocument
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			var AIProxyLibraryResponse AIProxyLibraryStreamResponse
+			err := json.Unmarshal([]byte(data), &AIProxyLibraryResponse)
+			if err != nil {
+				common.SysError("error unmarshalling stream response: " + err.Error())
+				return true
+			}
+			if len(AIProxyLibraryResponse.Documents) != 0 {
+				documents = AIProxyLibraryResponse.Documents
+			}
+			response := streamResponseAIProxyLibrary2OpenAI(&AIProxyLibraryResponse)
+			jsonResponse, err := json.Marshal(response)
+			if err != nil {
+				common.SysError("error marshalling stream response: " + err.Error())
+				return true
+			}
+			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
+			return true
+		case <-stopChan:
+			response := documentsAIProxyLibrary(documents)
+			jsonResponse, err := json.Marshal(response)
+			if err != nil {
+				common.SysError("error marshalling stream response: " + err.Error())
+				return true
+			}
+			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	return nil, &usage
+}
+
+func aiProxyLibraryHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var AIProxyLibraryResponse AIProxyLibraryResponse
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = json.Unmarshal(responseBody, &AIProxyLibraryResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if AIProxyLibraryResponse.ErrCode != 0 {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: AIProxyLibraryResponse.Message,
+				Type:    strconv.Itoa(AIProxyLibraryResponse.ErrCode),
+				Code:    AIProxyLibraryResponse.ErrCode,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseAIProxyLibrary2OpenAI(&AIProxyLibraryResponse)
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &fullTextResponse.Usage
+}

controller/relay-ali.go

@@ -35,6 +35,29 @@ type AliChatRequest struct {
 	Parameters AliParameters `json:"parameters,omitempty"`
 }
 
+type AliEmbeddingRequest struct {
+	Model string `json:"model"`
+	Input struct {
+		Texts []string `json:"texts"`
+	} `json:"input"`
+	Parameters *struct {
+		TextType string `json:"text_type,omitempty"`
+	} `json:"parameters,omitempty"`
+}
+
+type AliEmbedding struct {
+	Embedding []float64 `json:"embedding"`
+	TextIndex int       `json:"text_index"`
+}
+
+type AliEmbeddingResponse struct {
+	Output struct {
+		Embeddings []AliEmbedding `json:"embeddings"`
+	} `json:"output"`
+	Usage AliUsage `json:"usage"`
+	AliError
+}
+
 type AliError struct {
 	Code      string `json:"code"`
 	Message   string `json:"message"`
@@ -44,6 +67,7 @@ type AliError struct {
 type AliUsage struct {
 	InputTokens  int `json:"input_tokens"`
 	OutputTokens int `json:"output_tokens"`
+	TotalTokens  int `json:"total_tokens"`
 }
 
 type AliOutput struct {
@@ -64,18 +88,18 @@ func requestOpenAI2Ali(request GeneralOpenAIRequest) *AliChatRequest {
 		message := request.Messages[i]
 		if message.Role == "system" {
 			messages = append(messages, AliMessage{
-				User: message.Content,
+				User: message.StringContent(),
 				Bot:  "Okay",
 			})
 			continue
 		} else {
 			if i == len(request.Messages)-1 {
-				prompt = message.Content
+				prompt = message.StringContent()
 				break
 			}
 			messages = append(messages, AliMessage{
-				User: message.Content,
-				Bot:  request.Messages[i+1].Content,
+				User: message.StringContent(),
+				Bot:  request.Messages[i+1].StringContent(),
 			})
 			i++
 		}
@@ -95,6 +119,70 @@ func requestOpenAI2Ali(request GeneralOpenAIRequest) *AliChatRequest {
 	}
 }
 
+func embeddingRequestOpenAI2Ali(request GeneralOpenAIRequest) *AliEmbeddingRequest {
+	return &AliEmbeddingRequest{
+		Model: "text-embedding-v1",
+		Input: struct {
+			Texts []string `json:"texts"`
+		}{
+			Texts: request.ParseInput(),
+		},
+	}
+}
+
+func aliEmbeddingHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var aliResponse AliEmbeddingResponse
+	err := json.NewDecoder(resp.Body).Decode(&aliResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+
+	if aliResponse.Code != "" {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: aliResponse.Message,
+				Type:    aliResponse.Code,
+				Param:   aliResponse.RequestId,
+				Code:    aliResponse.Code,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+
+	fullTextResponse := embeddingResponseAli2OpenAI(&aliResponse)
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &fullTextResponse.Usage
+}
+
+func embeddingResponseAli2OpenAI(response *AliEmbeddingResponse) *OpenAIEmbeddingResponse {
+	openAIEmbeddingResponse := OpenAIEmbeddingResponse{
+		Object: "list",
+		Data:   make([]OpenAIEmbeddingResponseItem, 0, len(response.Output.Embeddings)),
+		Model:  "text-embedding-v1",
+		Usage:  Usage{TotalTokens: response.Usage.TotalTokens},
+	}
+
+	for _, item := range response.Output.Embeddings {
+		openAIEmbeddingResponse.Data = append(openAIEmbeddingResponse.Data, OpenAIEmbeddingResponseItem{
+			Object:    `embedding`,
+			Index:     item.TextIndex,
+			Embedding: item.Embedding,
+		})
+	}
+	return &openAIEmbeddingResponse
+}
+
 func responseAli2OpenAI(response *AliChatResponse) *OpenAITextResponse {
 	choice := OpenAITextResponseChoice{
 		Index: 0,

controller/relay-audio.go

@@ -0,0 +1,262 @@
+package controller
+
+import (
+	"bufio"
+	"bytes"
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"one-api/model"
+	"strings"
+)
+
+func relayAudioHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
+	audioModel := "whisper-1"
+
+	tokenId := c.GetInt("token_id")
+	channelType := c.GetInt("channel")
+	channelId := c.GetInt("channel_id")
+	userId := c.GetInt("id")
+	group := c.GetString("group")
+	tokenName := c.GetString("token_name")
+
+	var ttsRequest TextToSpeechRequest
+	if relayMode == RelayModeAudioSpeech {
+		// Read JSON
+		err := common.UnmarshalBodyReusable(c, &ttsRequest)
+		// Check if JSON is valid
+		if err != nil {
+			return errorWrapper(err, "invalid_json", http.StatusBadRequest)
+		}
+		audioModel = ttsRequest.Model
+		// Check if text is too long 4096
+		if len(ttsRequest.Input) > 4096 {
+			return errorWrapper(errors.New("input is too long (over 4096 characters)"), "text_too_long", http.StatusBadRequest)
+		}
+	}
+
+	modelRatio := common.GetModelRatio(audioModel)
+	groupRatio := common.GetGroupRatio(group)
+	ratio := modelRatio * groupRatio
+	var quota int
+	var preConsumedQuota int
+	switch relayMode {
+	case RelayModeAudioSpeech:
+		preConsumedQuota = int(float64(len(ttsRequest.Input)) * ratio)
+		quota = preConsumedQuota
+	default:
+		preConsumedQuota = int(float64(common.PreConsumedQuota) * ratio)
+	}
+	userQuota, err := model.CacheGetUserQuota(userId)
+	if err != nil {
+		return errorWrapper(err, "get_user_quota_failed", http.StatusInternalServerError)
+	}
+
+	// Check if user quota is enough
+	if userQuota-preConsumedQuota < 0 {
+		return errorWrapper(errors.New("user quota is not enough"), "insufficient_user_quota", http.StatusForbidden)
+	}
+	err = model.CacheDecreaseUserQuota(userId, preConsumedQuota)
+	if err != nil {
+		return errorWrapper(err, "decrease_user_quota_failed", http.StatusInternalServerError)
+	}
+	if userQuota > 100*preConsumedQuota {
+		// in this case, we do not pre-consume quota
+		// because the user has enough quota
+		preConsumedQuota = 0
+	}
+	if preConsumedQuota > 0 {
+		err := model.PreConsumeTokenQuota(tokenId, preConsumedQuota)
+		if err != nil {
+			return errorWrapper(err, "pre_consume_token_quota_failed", http.StatusForbidden)
+		}
+	}
+
+	// map model name
+	modelMapping := c.GetString("model_mapping")
+	if modelMapping != "" {
+		modelMap := make(map[string]string)
+		err := json.Unmarshal([]byte(modelMapping), &modelMap)
+		if err != nil {
+			return errorWrapper(err, "unmarshal_model_mapping_failed", http.StatusInternalServerError)
+		}
+		if modelMap[audioModel] != "" {
+			audioModel = modelMap[audioModel]
+		}
+	}
+
+	baseURL := common.ChannelBaseURLs[channelType]
+	requestURL := c.Request.URL.String()
+	if c.GetString("base_url") != "" {
+		baseURL = c.GetString("base_url")
+	}
+
+	fullRequestURL := getFullRequestURL(baseURL, requestURL, channelType)
+	if relayMode == RelayModeAudioTranscription && channelType == common.ChannelTypeAzure {
+		// https://learn.microsoft.com/en-us/azure/ai-services/openai/whisper-quickstart?tabs=command-line#rest-api
+		apiVersion := GetAPIVersion(c)
+		fullRequestURL = fmt.Sprintf("%s/openai/deployments/%s/audio/transcriptions?api-version=%s", baseURL, audioModel, apiVersion)
+	}
+
+	requestBody := &bytes.Buffer{}
+	_, err = io.Copy(requestBody, c.Request.Body)
+	if err != nil {
+		return errorWrapper(err, "new_request_body_failed", http.StatusInternalServerError)
+	}
+	c.Request.Body = io.NopCloser(bytes.NewBuffer(requestBody.Bytes()))
+	responseFormat := c.DefaultPostForm("response_format", "json")
+
+	req, err := http.NewRequest(c.Request.Method, fullRequestURL, requestBody)
+	if err != nil {
+		return errorWrapper(err, "new_request_failed", http.StatusInternalServerError)
+	}
+
+	if relayMode == RelayModeAudioTranscription && channelType == common.ChannelTypeAzure {
+		// https://learn.microsoft.com/en-us/azure/ai-services/openai/whisper-quickstart?tabs=command-line#rest-api
+		apiKey := c.Request.Header.Get("Authorization")
+		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+		req.Header.Set("api-key", apiKey)
+		req.ContentLength = c.Request.ContentLength
+	} else {
+		req.Header.Set("Authorization", c.Request.Header.Get("Authorization"))
+	}
+	req.Header.Set("Content-Type", c.Request.Header.Get("Content-Type"))
+	req.Header.Set("Accept", c.Request.Header.Get("Accept"))
+
+	resp, err := httpClient.Do(req)
+	if err != nil {
+		return errorWrapper(err, "do_request_failed", http.StatusInternalServerError)
+	}
+
+	err = req.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_request_body_failed", http.StatusInternalServerError)
+	}
+	err = c.Request.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_request_body_failed", http.StatusInternalServerError)
+	}
+
+	if relayMode != RelayModeAudioSpeech {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
+		}
+		err = resp.Body.Close()
+		if err != nil {
+			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
+		}
+
+		var openAIErr TextResponse
+		if err = json.Unmarshal(responseBody, &openAIErr); err == nil {
+			if openAIErr.Error.Message != "" {
+				return errorWrapper(fmt.Errorf("type %s, code %v, message %s", openAIErr.Error.Type, openAIErr.Error.Code, openAIErr.Error.Message), "request_error", http.StatusInternalServerError)
+			}
+		}
+
+		var text string
+		switch responseFormat {
+		case "json":
+			text, err = getTextFromJSON(responseBody)
+		case "text":
+			text, err = getTextFromText(responseBody)
+		case "srt":
+			text, err = getTextFromSRT(responseBody)
+		case "verbose_json":
+			text, err = getTextFromVerboseJSON(responseBody)
+		case "vtt":
+			text, err = getTextFromVTT(responseBody)
+		default:
+			return errorWrapper(errors.New("unexpected_response_format"), "unexpected_response_format", http.StatusInternalServerError)
+		}
+		if err != nil {
+			return errorWrapper(err, "get_text_from_body_err", http.StatusInternalServerError)
+		}
+		quota = countTokenText(text, audioModel)
+		resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	}
+	if resp.StatusCode != http.StatusOK {
+		if preConsumedQuota > 0 {
+			// we need to roll back the pre-consumed quota
+			defer func(ctx context.Context) {
+				go func() {
+					// negative means add quota back for token & user
+					err := model.PostConsumeTokenQuota(tokenId, -preConsumedQuota)
+					if err != nil {
+						common.LogError(ctx, fmt.Sprintf("error rollback pre-consumed quota: %s", err.Error()))
+					}
+				}()
+			}(c.Request.Context())
+		}
+		return relayErrorHandler(resp)
+	}
+	quotaDelta := quota - preConsumedQuota
+	defer func(ctx context.Context) {
+		go postConsumeQuota(ctx, tokenId, quotaDelta, quota, userId, channelId, modelRatio, groupRatio, audioModel, tokenName)
+	}(c.Request.Context())
+
+	for k, v := range resp.Header {
+		c.Writer.Header().Set(k, v[0])
+	}
+	c.Writer.WriteHeader(resp.StatusCode)
+
+	_, err = io.Copy(c.Writer, resp.Body)
+	if err != nil {
+		return errorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError)
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
+	}
+	return nil
+}
+
+func getTextFromVTT(body []byte) (string, error) {
+	return getTextFromSRT(body)
+}
+
+func getTextFromVerboseJSON(body []byte) (string, error) {
+	var whisperResponse WhisperVerboseJSONResponse
+	if err := json.Unmarshal(body, &whisperResponse); err != nil {
+		return "", fmt.Errorf("unmarshal_response_body_failed err :%w", err)
+	}
+	return whisperResponse.Text, nil
+}
+
+func getTextFromSRT(body []byte) (string, error) {
+	scanner := bufio.NewScanner(strings.NewReader(string(body)))
+	var builder strings.Builder
+	var textLine bool
+	for scanner.Scan() {
+		line := scanner.Text()
+		if textLine {
+			builder.WriteString(line)
+			textLine = false
+			continue
+		} else if strings.Contains(line, "-->") {
+			textLine = true
+			continue
+		}
+	}
+	if err := scanner.Err(); err != nil {
+		return "", err
+	}
+	return builder.String(), nil
+}
+
+func getTextFromText(body []byte) (string, error) {
+	return strings.TrimSuffix(string(body), "\n"), nil
+}
+
+func getTextFromJSON(body []byte) (string, error) {
+	var whisperResponse WhisperJSONResponse
+	if err := json.Unmarshal(body, &whisperResponse); err != nil {
+		return "", fmt.Errorf("unmarshal_response_body_failed err :%w", err)
+	}
+	return whisperResponse.Text, nil
+}

controller/relay-baidu.go

@@ -89,7 +89,7 @@ func requestOpenAI2Baidu(request GeneralOpenAIRequest) *BaiduChatRequest {
 		if message.Role == "system" {
 			messages = append(messages, BaiduMessage{
 				Role:    "user",
-				Content: message.Content,
+				Content: message.StringContent(),
 			})
 			messages = append(messages, BaiduMessage{
 				Role:    "assistant",
@@ -98,7 +98,7 @@ func requestOpenAI2Baidu(request GeneralOpenAIRequest) *BaiduChatRequest {
 		} else {
 			messages = append(messages, BaiduMessage{
 				Role:    message.Role,
-				Content: message.Content,
+				Content: message.StringContent(),
 			})
 		}
 	}
@@ -144,20 +144,9 @@ func streamResponseBaidu2OpenAI(baiduResponse *BaiduChatStreamResponse) *ChatCom
 }
 
 func embeddingRequestOpenAI2Baidu(request GeneralOpenAIRequest) *BaiduEmbeddingRequest {
-	baiduEmbeddingRequest := BaiduEmbeddingRequest{
-		Input: nil,
+	return &BaiduEmbeddingRequest{
+		Input: request.ParseInput(),
 	}
-	switch request.Input.(type) {
-	case string:
-		baiduEmbeddingRequest.Input = []string{request.Input.(string)}
-	case []any:
-		for _, item := range request.Input.([]any) {
-			if str, ok := item.(string); ok {
-				baiduEmbeddingRequest.Input = append(baiduEmbeddingRequest.Input, str)
-			}
-		}
-	}
-	return &baiduEmbeddingRequest
 }
 
 func embeddingResponseBaidu2OpenAI(response *BaiduEmbeddingResponse) *OpenAIEmbeddingResponse {

controller/relay-claude.go

@@ -70,7 +70,9 @@ func requestOpenAI2Claude(textRequest GeneralOpenAIRequest) *ClaudeRequest {
 		} else if message.Role == "assistant" {
 			prompt += fmt.Sprintf("\n\nAssistant: %s", message.Content)
 		} else if message.Role == "system" {
-			prompt += fmt.Sprintf("\n\nSystem: %s", message.Content)
+			if prompt == "" {
+				prompt = message.StringContent()
+			}
 		}
 	}
 	prompt += "\n\nAssistant:"

controller/relay-gemini.go

@@ -0,0 +1,281 @@
+package controller
+
+import (
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"one-api/common"
+
+	"github.com/gin-gonic/gin"
+)
+
+type GeminiChatRequest struct {
+	Contents         []GeminiChatContent        `json:"contents"`
+	SafetySettings   []GeminiChatSafetySettings `json:"safety_settings,omitempty"`
+	GenerationConfig GeminiChatGenerationConfig `json:"generation_config,omitempty"`
+	Tools            []GeminiChatTools          `json:"tools,omitempty"`
+}
+
+type GeminiInlineData struct {
+	MimeType string `json:"mimeType"`
+	Data     string `json:"data"`
+}
+
+type GeminiPart struct {
+	Text       string            `json:"text,omitempty"`
+	InlineData *GeminiInlineData `json:"inlineData,omitempty"`
+}
+
+type GeminiChatContent struct {
+	Role  string       `json:"role,omitempty"`
+	Parts []GeminiPart `json:"parts"`
+}
+
+type GeminiChatSafetySettings struct {
+	Category  string `json:"category"`
+	Threshold string `json:"threshold"`
+}
+
+type GeminiChatTools struct {
+	FunctionDeclarations any `json:"functionDeclarations,omitempty"`
+}
+
+type GeminiChatGenerationConfig struct {
+	Temperature     float64  `json:"temperature,omitempty"`
+	TopP            float64  `json:"topP,omitempty"`
+	TopK            float64  `json:"topK,omitempty"`
+	MaxOutputTokens int      `json:"maxOutputTokens,omitempty"`
+	CandidateCount  int      `json:"candidateCount,omitempty"`
+	StopSequences   []string `json:"stopSequences,omitempty"`
+}
+
+// Setting safety to the lowest possible values since Gemini is already powerless enough
+func requestOpenAI2Gemini(textRequest GeneralOpenAIRequest) *GeminiChatRequest {
+	geminiRequest := GeminiChatRequest{
+		Contents: make([]GeminiChatContent, 0, len(textRequest.Messages)),
+		//SafetySettings: []GeminiChatSafetySettings{
+		//	{
+		//		Category:  "HARM_CATEGORY_HARASSMENT",
+		//		Threshold: "BLOCK_ONLY_HIGH",
+		//	},
+		//	{
+		//		Category:  "HARM_CATEGORY_HATE_SPEECH",
+		//		Threshold: "BLOCK_ONLY_HIGH",
+		//	},
+		//	{
+		//		Category:  "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+		//		Threshold: "BLOCK_ONLY_HIGH",
+		//	},
+		//	{
+		//		Category:  "HARM_CATEGORY_DANGEROUS_CONTENT",
+		//		Threshold: "BLOCK_ONLY_HIGH",
+		//	},
+		//},
+		GenerationConfig: GeminiChatGenerationConfig{
+			Temperature:     textRequest.Temperature,
+			TopP:            textRequest.TopP,
+			MaxOutputTokens: textRequest.MaxTokens,
+		},
+	}
+	if textRequest.Functions != nil {
+		geminiRequest.Tools = []GeminiChatTools{
+			{
+				FunctionDeclarations: textRequest.Functions,
+			},
+		}
+	}
+	shouldAddDummyModelMessage := false
+	for _, message := range textRequest.Messages {
+		content := GeminiChatContent{
+			Role: message.Role,
+			Parts: []GeminiPart{
+				{
+					Text: message.StringContent(),
+				},
+			},
+		}
+		// there's no assistant role in gemini and API shall vomit if Role is not user or model
+		if content.Role == "assistant" {
+			content.Role = "model"
+		}
+		// Converting system prompt to prompt from user for the same reason
+		if content.Role == "system" {
+			content.Role = "user"
+			shouldAddDummyModelMessage = true
+		}
+		geminiRequest.Contents = append(geminiRequest.Contents, content)
+
+		// If a system message is the last message, we need to add a dummy model message to make gemini happy
+		if shouldAddDummyModelMessage {
+			geminiRequest.Contents = append(geminiRequest.Contents, GeminiChatContent{
+				Role: "model",
+				Parts: []GeminiPart{
+					{
+						Text: "ok",
+					},
+				},
+			})
+			shouldAddDummyModelMessage = false
+		}
+	}
+
+	return &geminiRequest
+}
+
+type GeminiChatResponse struct {
+	Candidates     []GeminiChatCandidate    `json:"candidates"`
+	PromptFeedback GeminiChatPromptFeedback `json:"promptFeedback"`
+}
+
+type GeminiChatCandidate struct {
+	Content       GeminiChatContent        `json:"content"`
+	FinishReason  string                   `json:"finishReason"`
+	Index         int64                    `json:"index"`
+	SafetyRatings []GeminiChatSafetyRating `json:"safetyRatings"`
+}
+
+type GeminiChatSafetyRating struct {
+	Category    string `json:"category"`
+	Probability string `json:"probability"`
+}
+
+type GeminiChatPromptFeedback struct {
+	SafetyRatings []GeminiChatSafetyRating `json:"safetyRatings"`
+}
+
+func responseGeminiChat2OpenAI(response *GeminiChatResponse) *OpenAITextResponse {
+	fullTextResponse := OpenAITextResponse{
+		Id:      fmt.Sprintf("chatcmpl-%s", common.GetUUID()),
+		Object:  "chat.completion",
+		Created: common.GetTimestamp(),
+		Choices: make([]OpenAITextResponseChoice, 0, len(response.Candidates)),
+	}
+	for i, candidate := range response.Candidates {
+		choice := OpenAITextResponseChoice{
+			Index: i,
+			Message: Message{
+				Role:    "assistant",
+				Content: candidate.Content.Parts[0].Text,
+			},
+			FinishReason: stopFinishReason,
+		}
+		fullTextResponse.Choices = append(fullTextResponse.Choices, choice)
+	}
+	return &fullTextResponse
+}
+
+func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) *ChatCompletionsStreamResponse {
+	var choice ChatCompletionsStreamResponseChoice
+	if len(geminiResponse.Candidates) > 0 && len(geminiResponse.Candidates[0].Content.Parts) > 0 {
+		choice.Delta.Content = geminiResponse.Candidates[0].Content.Parts[0].Text
+	}
+	choice.FinishReason = &stopFinishReason
+	var response ChatCompletionsStreamResponse
+	response.Object = "chat.completion.chunk"
+	response.Model = "gemini"
+	response.Choices = []ChatCompletionsStreamResponseChoice{choice}
+	return &response
+}
+
+func geminiChatStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, string) {
+	responseText := ""
+	responseId := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
+	createdTime := common.GetTimestamp()
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			common.SysError("error reading stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		err = resp.Body.Close()
+		if err != nil {
+			common.SysError("error closing stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		var geminiResponse GeminiChatResponse
+		err = json.Unmarshal(responseBody, &geminiResponse)
+		if err != nil {
+			common.SysError("error unmarshalling stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		fullTextResponse := streamResponseGeminiChat2OpenAI(&geminiResponse)
+		fullTextResponse.Id = responseId
+		fullTextResponse.Created = createdTime
+		if len(geminiResponse.Candidates) > 0 && len(geminiResponse.Candidates[0].Content.Parts) > 0 {
+			responseText += geminiResponse.Candidates[0].Content.Parts[0].Text
+		}
+		jsonResponse, err := json.Marshal(fullTextResponse)
+		if err != nil {
+			common.SysError("error marshalling stream response: " + err.Error())
+			stopChan <- true
+			return
+		}
+		dataChan <- string(jsonResponse)
+		stopChan <- true
+	}()
+	setEventStreamHeaders(c)
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			c.Render(-1, common.CustomEvent{Data: "data: " + data})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func geminiChatHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*OpenAIErrorWithStatusCode, *Usage) {
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	var geminiResponse GeminiChatResponse
+	err = json.Unmarshal(responseBody, &geminiResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if len(geminiResponse.Candidates) == 0 {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: "No candidates returned",
+				Type:    "server_error",
+				Param:   "",
+				Code:    500,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseGeminiChat2OpenAI(&geminiResponse)
+	completionTokens := countTokenText(geminiResponse.Candidates[0].Content.Parts[0].Text, model)
+	usage := Usage{
+		PromptTokens:     promptTokens,
+		CompletionTokens: completionTokens,
+		TotalTokens:      promptTokens + completionTokens,
+	}
+	fullTextResponse.Usage = usage
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &usage
+}

controller/relay-image.go

@@ -2,6 +2,7 @@ package controller
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -9,40 +10,76 @@ import (
 	"net/http"
 	"one-api/common"
 	"one-api/model"
+	"strings"
 
 	"github.com/gin-gonic/gin"
 )
 
+func isWithinRange(element string, value int) bool {
+	if _, ok := common.DalleGenerationImageAmounts[element]; !ok {
+		return false
+	}
+
+	min := common.DalleGenerationImageAmounts[element][0]
+	max := common.DalleGenerationImageAmounts[element][1]
+
+	return value >= min && value <= max
+}
+
 func relayImageHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
-	imageModel := "dall-e"
+	imageModel := "dall-e-2"
+	imageSize := "1024x1024"
 
 	tokenId := c.GetInt("token_id")
 	channelType := c.GetInt("channel")
+	channelId := c.GetInt("channel_id")
 	userId := c.GetInt("id")
-	consumeQuota := c.GetBool("consume_quota")
 	group := c.GetString("group")
 
 	var imageRequest ImageRequest
-	if consumeQuota {
-		err := common.UnmarshalBodyReusable(c, &imageRequest)
-		if err != nil {
-			return errorWrapper(err, "bind_request_body_failed", http.StatusBadRequest)
+	err := common.UnmarshalBodyReusable(c, &imageRequest)
+	if err != nil {
+		return errorWrapper(err, "bind_request_body_failed", http.StatusBadRequest)
+	}
+
+	// Size validation
+	if imageRequest.Size != "" {
+		imageSize = imageRequest.Size
+	}
+
+	// Model validation
+	if imageRequest.Model != "" {
+		imageModel = imageRequest.Model
+	}
+
+	imageCostRatio, hasValidSize := common.DalleSizeRatios[imageModel][imageSize]
+
+	// Check if model is supported
+	if hasValidSize {
+		if imageRequest.Quality == "hd" && imageModel == "dall-e-3" {
+			if imageSize == "1024x1024" {
+				imageCostRatio *= 2
+			} else {
+				imageCostRatio *= 1.5
+			}
 		}
+	} else {
+		return errorWrapper(errors.New("size not supported for this image model"), "size_not_supported", http.StatusBadRequest)
 	}
 
 	// Prompt validation
 	if imageRequest.Prompt == "" {
-		return errorWrapper(errors.New("prompt is required"), "required_field_missing", http.StatusBadRequest)
+		return errorWrapper(errors.New("prompt is required"), "prompt_missing", http.StatusBadRequest)
 	}
 
-	// Not "256x256", "512x512", or "1024x1024"
-	if imageRequest.Size != "" && imageRequest.Size != "256x256" && imageRequest.Size != "512x512" && imageRequest.Size != "1024x1024" {
-		return errorWrapper(errors.New("size must be one of 256x256, 512x512, or 1024x1024"), "invalid_field_value", http.StatusBadRequest)
+	// Check prompt length
+	if len(imageRequest.Prompt) > common.DalleImagePromptLengthLimitations[imageModel] {
+		return errorWrapper(errors.New("prompt is too long"), "prompt_too_long", http.StatusBadRequest)
 	}
 
-	// N should between 1 and 10
-	if imageRequest.N != 0 && (imageRequest.N < 1 || imageRequest.N > 10) {
-		return errorWrapper(errors.New("n must be between 1 and 10"), "invalid_field_value", http.StatusBadRequest)
+	// Number of generated images validation
+	if isWithinRange(imageModel, imageRequest.N) == false {
+		return errorWrapper(errors.New("invalid value of n"), "n_not_within_range", http.StatusBadRequest)
 	}
 
 	// map model name
@@ -59,18 +96,21 @@ func relayImageHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode
 			isModelMapped = true
 		}
 	}
-
 	baseURL := common.ChannelBaseURLs[channelType]
 	requestURL := c.Request.URL.String()
-
 	if c.GetString("base_url") != "" {
 		baseURL = c.GetString("base_url")
 	}
-
-	fullRequestURL := fmt.Sprintf("%s%s", baseURL, requestURL)
+	fullRequestURL := getFullRequestURL(baseURL, requestURL, channelType)
+	if channelType == common.ChannelTypeAzure && relayMode == RelayModeImagesGenerations {
+		// https://learn.microsoft.com/en-us/azure/ai-services/openai/dall-e-quickstart?tabs=dalle3%2Ccommand-line&pivots=rest-api
+		apiVersion := GetAPIVersion(c)
+		// https://{resource_name}.openai.azure.com/openai/deployments/dall-e-3/images/generations?api-version=2023-06-01-preview
+		fullRequestURL = fmt.Sprintf("%s/openai/deployments/%s/images/generations?api-version=%s", baseURL, imageModel, apiVersion)
+	}
 
 	var requestBody io.Reader
-	if isModelMapped {
+	if isModelMapped || channelType == common.ChannelTypeAzure { // make Azure channel request body
 		jsonStr, err := json.Marshal(imageRequest)
 		if err != nil {
 			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
@@ -85,26 +125,23 @@ func relayImageHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode
 	ratio := modelRatio * groupRatio
 	userQuota, err := model.CacheGetUserQuota(userId)
 
-	sizeRatio := 1.0
-	// Size
-	if imageRequest.Size == "256x256" {
-		sizeRatio = 1
-	} else if imageRequest.Size == "512x512" {
-		sizeRatio = 1.125
-	} else if imageRequest.Size == "1024x1024" {
-		sizeRatio = 1.25
-	}
-	quota := int(ratio*sizeRatio*1000) * imageRequest.N
+	quota := int(ratio*imageCostRatio*1000) * imageRequest.N
 
-	if consumeQuota && userQuota-quota < 0 {
-		return errorWrapper(err, "insufficient_user_quota", http.StatusForbidden)
+	if userQuota-quota < 0 {
+		return errorWrapper(errors.New("user quota is not enough"), "insufficient_user_quota", http.StatusForbidden)
 	}
 
 	req, err := http.NewRequest(c.Request.Method, fullRequestURL, requestBody)
 	if err != nil {
 		return errorWrapper(err, "new_request_failed", http.StatusInternalServerError)
 	}
-	req.Header.Set("Authorization", c.Request.Header.Get("Authorization"))
+	token := c.Request.Header.Get("Authorization")
+	if channelType == common.ChannelTypeAzure { // Azure authentication
+		token = strings.TrimPrefix(token, "Bearer ")
+		req.Header.Set("api-key", token)
+	} else {
+		req.Header.Set("Authorization", token)
+	}
 
 	req.Header.Set("Content-Type", c.Request.Header.Get("Content-Type"))
 	req.Header.Set("Accept", c.Request.Header.Get("Accept"))
@@ -124,44 +161,40 @@ func relayImageHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode
 	}
 	var textResponse ImageResponse
 
-	defer func() {
-		if consumeQuota {
-			err := model.PostConsumeTokenQuota(tokenId, quota)
-			if err != nil {
-				common.SysError("error consuming token remain quota: " + err.Error())
-			}
-			err = model.CacheUpdateUserQuota(userId)
-			if err != nil {
-				common.SysError("error update user quota cache: " + err.Error())
-			}
-			if quota != 0 {
-				tokenName := c.GetString("token_name")
-				logContent := fmt.Sprintf("模型倍率 %.2f，分组倍率 %.2f", modelRatio, groupRatio)
-				model.RecordConsumeLog(userId, 0, 0, imageModel, tokenName, quota, logContent)
-				model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
-				channelId := c.GetInt("channel_id")
-				model.UpdateChannelUsedQuota(channelId, quota)
-			}
-		}
-	}()
-
-	if consumeQuota {
-		responseBody, err := io.ReadAll(resp.Body)
-
+	defer func(ctx context.Context) {
+		err := model.PostConsumeTokenQuota(tokenId, quota)
 		if err != nil {
-			return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
+			common.SysError("error consuming token remain quota: " + err.Error())
 		}
-		err = resp.Body.Close()
+		err = model.CacheUpdateUserQuota(userId)
 		if err != nil {
-			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
+			common.SysError("error update user quota cache: " + err.Error())
 		}
-		err = json.Unmarshal(responseBody, &textResponse)
-		if err != nil {
-			return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError)
+		if quota != 0 {
+			tokenName := c.GetString("token_name")
+			logContent := fmt.Sprintf("模型倍率 %.2f，分组倍率 %.2f", modelRatio, groupRatio)
+			model.RecordConsumeLog(ctx, userId, channelId, 0, 0, imageModel, tokenName, quota, logContent)
+			model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
+			channelId := c.GetInt("channel_id")
+			model.UpdateChannelUsedQuota(channelId, quota)
 		}
+	}(c.Request.Context())
 
-		resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	responseBody, err := io.ReadAll(resp.Body)
+
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError)
 	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError)
+	}
+	err = json.Unmarshal(responseBody, &textResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError)
+	}
+
+	resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
 
 	for k, v := range resp.Header {
 		c.Writer.Header().Set(k, v[0])

controller/relay-openai.go

@@ -88,30 +88,29 @@ func openaiStreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*O
 	return nil, responseText
 }
 
-func openaiHandler(c *gin.Context, resp *http.Response, consumeQuota bool, promptTokens int, model string) (*OpenAIErrorWithStatusCode, *Usage) {
+func openaiHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*OpenAIErrorWithStatusCode, *Usage) {
 	var textResponse TextResponse
-	if consumeQuota {
-		responseBody, err := io.ReadAll(resp.Body)
-		if err != nil {
-			return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
-		}
-		err = resp.Body.Close()
-		if err != nil {
-			return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
-		}
-		err = json.Unmarshal(responseBody, &textResponse)
-		if err != nil {
-			return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
-		}
-		if textResponse.Error.Type != "" {
-			return &OpenAIErrorWithStatusCode{
-				OpenAIError: textResponse.Error,
-				StatusCode:  resp.StatusCode,
-			}, nil
-		}
-		// Reset response body
-		resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
 	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = json.Unmarshal(responseBody, &textResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if textResponse.Error.Type != "" {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: textResponse.Error,
+			StatusCode:  resp.StatusCode,
+		}, nil
+	}
+	// Reset response body
+	resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+
 	// We shouldn't set the header before we parse the response body, because the parse part may fail.
 	// And then we will have to send an error response, but in this case, the header has already been set.
 	// So the httpClient will be confused by the response.
@@ -120,7 +119,7 @@ func openaiHandler(c *gin.Context, resp *http.Response, consumeQuota bool, promp
 		c.Writer.Header().Set(k, v[0])
 	}
 	c.Writer.WriteHeader(resp.StatusCode)
-	_, err := io.Copy(c.Writer, resp.Body)
+	_, err = io.Copy(c.Writer, resp.Body)
 	if err != nil {
 		return errorWrapper(err, "copy_response_body_failed", http.StatusInternalServerError), nil
 	}
@@ -132,7 +131,7 @@ func openaiHandler(c *gin.Context, resp *http.Response, consumeQuota bool, promp
 	if textResponse.Usage.TotalTokens == 0 {
 		completionTokens := 0
 		for _, choice := range textResponse.Choices {
-			completionTokens += countTokenText(choice.Message.Content, model)
+			completionTokens += countTokenText(choice.Message.StringContent(), model)
 		}
 		textResponse.Usage = Usage{
 			PromptTokens:     promptTokens,

controller/relay-palm.go

@@ -59,7 +59,7 @@ func requestOpenAI2PaLM(textRequest GeneralOpenAIRequest) *PaLMChatRequest {
 	}
 	for _, message := range textRequest.Messages {
 		palmMessage := PaLMChatMessage{
-			Content: message.Content,
+			Content: message.StringContent(),
 		}
 		if message.Role == "user" {
 			palmMessage.Author = "0"

controller/relay-tencent.go

@@ -0,0 +1,287 @@
+package controller
+
+import (
+	"bufio"
+	"crypto/hmac"
+	"crypto/sha1"
+	"encoding/base64"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"io"
+	"net/http"
+	"one-api/common"
+	"sort"
+	"strconv"
+	"strings"
+)
+
+// https://cloud.tencent.com/document/product/1729/97732
+
+type TencentMessage struct {
+	Role    string `json:"role"`
+	Content string `json:"content"`
+}
+
+type TencentChatRequest struct {
+	AppId    int64  `json:"app_id"`    // 腾讯云账号的 APPID
+	SecretId string `json:"secret_id"` // 官网 SecretId
+	// Timestamp当前 UNIX 时间戳，单位为秒，可记录发起 API 请求的时间。
+	// 例如1529223702，如果与当前时间相差过大，会引起签名过期错误
+	Timestamp int64 `json:"timestamp"`
+	// Expired 签名的有效期，是一个符合 UNIX Epoch 时间戳规范的数值，
+	// 单位为秒；Expired 必须大于 Timestamp 且 Expired-Timestamp 小于90天
+	Expired int64  `json:"expired"`
+	QueryID string `json:"query_id"` //请求 Id，用于问题排查
+	// Temperature 较高的数值会使输出更加随机，而较低的数值会使其更加集中和确定
+	// 默认 1.0，取值区间为[0.0,2.0]，非必要不建议使用,不合理的取值会影响效果
+	// 建议该参数和 top_p 只设置1个，不要同时更改 top_p
+	Temperature float64 `json:"temperature"`
+	// TopP 影响输出文本的多样性，取值越大，生成文本的多样性越强
+	// 默认1.0，取值区间为[0.0, 1.0]，非必要不建议使用, 不合理的取值会影响效果
+	// 建议该参数和 temperature 只设置1个，不要同时更改
+	TopP float64 `json:"top_p"`
+	// Stream 0：同步，1：流式 （默认，协议：SSE)
+	// 同步请求超时：60s，如果内容较长建议使用流式
+	Stream int `json:"stream"`
+	// Messages 会话内容, 长度最多为40, 按对话时间从旧到新在数组中排列
+	// 输入 content 总数最大支持 3000 token。
+	Messages []TencentMessage `json:"messages"`
+}
+
+type TencentError struct {
+	Code    int    `json:"code"`
+	Message string `json:"message"`
+}
+
+type TencentUsage struct {
+	InputTokens  int `json:"input_tokens"`
+	OutputTokens int `json:"output_tokens"`
+	TotalTokens  int `json:"total_tokens"`
+}
+
+type TencentResponseChoices struct {
+	FinishReason string         `json:"finish_reason,omitempty"` // 流式结束标志位，为 stop 则表示尾包
+	Messages     TencentMessage `json:"messages,omitempty"`      // 内容，同步模式返回内容，流模式为 null 输出 content 内容总数最多支持 1024token。
+	Delta        TencentMessage `json:"delta,omitempty"`         // 内容，流模式返回内容，同步模式为 null 输出 content 内容总数最多支持 1024token。
+}
+
+type TencentChatResponse struct {
+	Choices []TencentResponseChoices `json:"choices,omitempty"` // 结果
+	Created string                   `json:"created,omitempty"` // unix 时间戳的字符串
+	Id      string                   `json:"id,omitempty"`      // 会话 id
+	Usage   Usage                    `json:"usage,omitempty"`   // token 数量
+	Error   TencentError             `json:"error,omitempty"`   // 错误信息 注意：此字段可能返回 null，表示取不到有效值
+	Note    string                   `json:"note,omitempty"`    // 注释
+	ReqID   string                   `json:"req_id,omitempty"`  // 唯一请求 Id，每次请求都会返回。用于反馈接口入参
+}
+
+func requestOpenAI2Tencent(request GeneralOpenAIRequest) *TencentChatRequest {
+	messages := make([]TencentMessage, 0, len(request.Messages))
+	for i := 0; i < len(request.Messages); i++ {
+		message := request.Messages[i]
+		if message.Role == "system" {
+			messages = append(messages, TencentMessage{
+				Role:    "user",
+				Content: message.StringContent(),
+			})
+			messages = append(messages, TencentMessage{
+				Role:    "assistant",
+				Content: "Okay",
+			})
+			continue
+		}
+		messages = append(messages, TencentMessage{
+			Content: message.StringContent(),
+			Role:    message.Role,
+		})
+	}
+	stream := 0
+	if request.Stream {
+		stream = 1
+	}
+	return &TencentChatRequest{
+		Timestamp:   common.GetTimestamp(),
+		Expired:     common.GetTimestamp() + 24*60*60,
+		QueryID:     common.GetUUID(),
+		Temperature: request.Temperature,
+		TopP:        request.TopP,
+		Stream:      stream,
+		Messages:    messages,
+	}
+}
+
+func responseTencent2OpenAI(response *TencentChatResponse) *OpenAITextResponse {
+	fullTextResponse := OpenAITextResponse{
+		Object:  "chat.completion",
+		Created: common.GetTimestamp(),
+		Usage:   response.Usage,
+	}
+	if len(response.Choices) > 0 {
+		choice := OpenAITextResponseChoice{
+			Index: 0,
+			Message: Message{
+				Role:    "assistant",
+				Content: response.Choices[0].Messages.Content,
+			},
+			FinishReason: response.Choices[0].FinishReason,
+		}
+		fullTextResponse.Choices = append(fullTextResponse.Choices, choice)
+	}
+	return &fullTextResponse
+}
+
+func streamResponseTencent2OpenAI(TencentResponse *TencentChatResponse) *ChatCompletionsStreamResponse {
+	response := ChatCompletionsStreamResponse{
+		Object:  "chat.completion.chunk",
+		Created: common.GetTimestamp(),
+		Model:   "tencent-hunyuan",
+	}
+	if len(TencentResponse.Choices) > 0 {
+		var choice ChatCompletionsStreamResponseChoice
+		choice.Delta.Content = TencentResponse.Choices[0].Delta.Content
+		if TencentResponse.Choices[0].FinishReason == "stop" {
+			choice.FinishReason = &stopFinishReason
+		}
+		response.Choices = append(response.Choices, choice)
+	}
+	return &response
+}
+
+func tencentStreamHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, string) {
+	var responseText string
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
+		if atEOF && len(data) == 0 {
+			return 0, nil, nil
+		}
+		if i := strings.Index(string(data), "\n"); i >= 0 {
+			return i + 1, data[0:i], nil
+		}
+		if atEOF {
+			return len(data), data, nil
+		}
+		return 0, nil, nil
+	})
+	dataChan := make(chan string)
+	stopChan := make(chan bool)
+	go func() {
+		for scanner.Scan() {
+			data := scanner.Text()
+			if len(data) < 5 { // ignore blank line or wrong format
+				continue
+			}
+			if data[:5] != "data:" {
+				continue
+			}
+			data = data[5:]
+			dataChan <- data
+		}
+		stopChan <- true
+	}()
+	setEventStreamHeaders(c)
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case data := <-dataChan:
+			var TencentResponse TencentChatResponse
+			err := json.Unmarshal([]byte(data), &TencentResponse)
+			if err != nil {
+				common.SysError("error unmarshalling stream response: " + err.Error())
+				return true
+			}
+			response := streamResponseTencent2OpenAI(&TencentResponse)
+			if len(response.Choices) != 0 {
+				responseText += response.Choices[0].Delta.Content
+			}
+			jsonResponse, err := json.Marshal(response)
+			if err != nil {
+				common.SysError("error marshalling stream response: " + err.Error())
+				return true
+			}
+			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	err := resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), ""
+	}
+	return nil, responseText
+}
+
+func tencentHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
+	var TencentResponse TencentChatResponse
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = json.Unmarshal(responseBody, &TencentResponse)
+	if err != nil {
+		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if TencentResponse.Error.Code != 0 {
+		return &OpenAIErrorWithStatusCode{
+			OpenAIError: OpenAIError{
+				Message: TencentResponse.Error.Message,
+				Code:    TencentResponse.Error.Code,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+	fullTextResponse := responseTencent2OpenAI(&TencentResponse)
+	jsonResponse, err := json.Marshal(fullTextResponse)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	c.Writer.Header().Set("Content-Type", "application/json")
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = c.Writer.Write(jsonResponse)
+	return nil, &fullTextResponse.Usage
+}
+
+func parseTencentConfig(config string) (appId int64, secretId string, secretKey string, err error) {
+	parts := strings.Split(config, "|")
+	if len(parts) != 3 {
+		err = errors.New("invalid tencent config")
+		return
+	}
+	appId, err = strconv.ParseInt(parts[0], 10, 64)
+	secretId = parts[1]
+	secretKey = parts[2]
+	return
+}
+
+func getTencentSign(req TencentChatRequest, secretKey string) string {
+	params := make([]string, 0)
+	params = append(params, "app_id="+strconv.FormatInt(req.AppId, 10))
+	params = append(params, "secret_id="+req.SecretId)
+	params = append(params, "timestamp="+strconv.FormatInt(req.Timestamp, 10))
+	params = append(params, "query_id="+req.QueryID)
+	params = append(params, "temperature="+strconv.FormatFloat(req.Temperature, 'f', -1, 64))
+	params = append(params, "top_p="+strconv.FormatFloat(req.TopP, 'f', -1, 64))
+	params = append(params, "stream="+strconv.Itoa(req.Stream))
+	params = append(params, "expired="+strconv.FormatInt(req.Expired, 10))
+
+	var messageStr string
+	for _, msg := range req.Messages {
+		messageStr += fmt.Sprintf(`{"role":"%s","content":"%s"},`, msg.Role, msg.Content)
+	}
+	messageStr = strings.TrimSuffix(messageStr, ",")
+	params = append(params, "messages=["+messageStr+"]")
+
+	sort.Sort(sort.StringSlice(params))
+	url := "hunyuan.cloud.tencent.com/hyllm/v1/chat/completions?" + strings.Join(params, "&")
+	mac := hmac.New(sha1.New, []byte(secretKey))
+	signURL := url
+	mac.Write([]byte(signURL))
+	sign := mac.Sum([]byte(nil))
+	return base64.StdEncoding.EncodeToString(sign)
+}

controller/relay-text.go

@@ -2,16 +2,19 @@ package controller
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
+	"math"
 	"net/http"
 	"one-api/common"
 	"one-api/model"
 	"strings"
 	"time"
+
+	"github.com/gin-gonic/gin"
 )
 
 const (
@@ -22,13 +25,23 @@ const (
 	APITypeZhipu
 	APITypeAli
 	APITypeXunfei
+	APITypeAIProxyLibrary
+	APITypeTencent
+	APITypeGemini
 )
 
 var httpClient *http.Client
 var impatientHTTPClient *http.Client
 
 func init() {
-	httpClient = &http.Client{}
+	if common.RelayTimeout == 0 {
+		httpClient = &http.Client{}
+	} else {
+		httpClient = &http.Client{
+			Timeout: time.Duration(common.RelayTimeout) * time.Second,
+		}
+	}
+
 	impatientHTTPClient = &http.Client{
 		Timeout: 5 * time.Second,
 	}
@@ -36,16 +49,14 @@ func init() {
 
 func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 	channelType := c.GetInt("channel")
+	channelId := c.GetInt("channel_id")
 	tokenId := c.GetInt("token_id")
 	userId := c.GetInt("id")
-	consumeQuota := c.GetBool("consume_quota")
 	group := c.GetString("group")
 	var textRequest GeneralOpenAIRequest
-	if consumeQuota || channelType == common.ChannelTypeAzure || channelType == common.ChannelTypePaLM {
-		err := common.UnmarshalBodyReusable(c, &textRequest)
-		if err != nil {
-			return errorWrapper(err, "bind_request_body_failed", http.StatusBadRequest)
-		}
+	err := common.UnmarshalBodyReusable(c, &textRequest)
+	if err != nil {
+		return errorWrapper(err, "bind_request_body_failed", http.StatusBadRequest)
 	}
 	if relayMode == RelayModeModerations && textRequest.Model == "" {
 		textRequest.Model = "text-moderation-latest"
@@ -104,22 +115,24 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 		apiType = APITypeAli
 	case common.ChannelTypeXunfei:
 		apiType = APITypeXunfei
+	case common.ChannelTypeAIProxyLibrary:
+		apiType = APITypeAIProxyLibrary
+	case common.ChannelTypeTencent:
+		apiType = APITypeTencent
+	case common.ChannelTypeGemini:
+		apiType = APITypeGemini
 	}
 	baseURL := common.ChannelBaseURLs[channelType]
 	requestURL := c.Request.URL.String()
 	if c.GetString("base_url") != "" {
 		baseURL = c.GetString("base_url")
 	}
-	fullRequestURL := fmt.Sprintf("%s%s", baseURL, requestURL)
+	fullRequestURL := getFullRequestURL(baseURL, requestURL, channelType)
 	switch apiType {
 	case APITypeOpenAI:
 		if channelType == common.ChannelTypeAzure {
 			// https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?pivots=rest-api&tabs=command-line#rest-api
-			query := c.Request.URL.Query()
-			apiVersion := query.Get("api-version")
-			if apiVersion == "" {
-				apiVersion = c.GetString("api_version")
-			}
+			apiVersion := GetAPIVersion(c)
 			requestURL := strings.Split(requestURL, "?")[0]
 			requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, apiVersion)
 			baseURL = c.GetString("base_url")
@@ -130,7 +143,9 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			model_ = strings.TrimSuffix(model_, "-0301")
 			model_ = strings.TrimSuffix(model_, "-0314")
 			model_ = strings.TrimSuffix(model_, "-0613")
-			fullRequestURL = fmt.Sprintf("%s/openai/deployments/%s/%s", baseURL, model_, task)
+
+			requestURL = fmt.Sprintf("/openai/deployments/%s/%s", model_, task)
+			fullRequestURL = getFullRequestURL(baseURL, requestURL, channelType)
 		}
 	case APITypeClaude:
 		fullRequestURL = "https://api.anthropic.com/v1/complete"
@@ -143,6 +158,8 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions"
 		case "ERNIE-Bot-turbo":
 			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant"
+		case "ERNIE-Bot-4":
+			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/completions_pro"
 		case "BLOOMZ-7B":
 			fullRequestURL = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/bloomz_7b1"
 		case "Embedding-V1":
@@ -163,6 +180,24 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 		apiKey := c.Request.Header.Get("Authorization")
 		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
 		fullRequestURL += "?key=" + apiKey
+	case APITypeGemini:
+		requestBaseURL := "https://generativelanguage.googleapis.com"
+		if baseURL != "" {
+			requestBaseURL = baseURL
+		}
+		version := "v1"
+		if c.GetString("api_version") != "" {
+			version = c.GetString("api_version")
+		}
+		action := "generateContent"
+		// actually gemini does not support stream, it's fake
+		//if textRequest.Stream {
+		//	action = "streamGenerateContent"
+		//}
+		fullRequestURL = fmt.Sprintf("%s/%s/models/%s:%s", requestBaseURL, version, textRequest.Model, action)
+		apiKey := c.Request.Header.Get("Authorization")
+		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+		fullRequestURL += "?key=" + apiKey
 	case APITypeZhipu:
 		method := "invoke"
 		if textRequest.Stream {
@@ -171,6 +206,13 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 		fullRequestURL = fmt.Sprintf("https://open.bigmodel.cn/api/paas/v3/model-api/%s/%s", textRequest.Model, method)
 	case APITypeAli:
 		fullRequestURL = "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation"
+		if relayMode == RelayModeEmbeddings {
+			fullRequestURL = "https://dashscope.aliyuncs.com/api/v1/services/embeddings/text-embedding/text-embedding"
+		}
+	case APITypeTencent:
+		fullRequestURL = "https://hunyuan.cloud.tencent.com/hyllm/v1/chat/completions"
+	case APITypeAIProxyLibrary:
+		fullRequestURL = fmt.Sprintf("%s/api/library/ask", baseURL)
 	}
 	var promptTokens int
 	var completionTokens int
@@ -194,6 +236,9 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 	if err != nil {
 		return errorWrapper(err, "get_user_quota_failed", http.StatusInternalServerError)
 	}
+	if userQuota-preConsumedQuota < 0 {
+		return errorWrapper(errors.New("user quota is not enough"), "insufficient_user_quota", http.StatusForbidden)
+	}
 	err = model.CacheDecreaseUserQuota(userId, preConsumedQuota)
 	if err != nil {
 		return errorWrapper(err, "decrease_user_quota_failed", http.StatusInternalServerError)
@@ -202,8 +247,9 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 		// in this case, we do not pre-consume quota
 		// because the user has enough quota
 		preConsumedQuota = 0
+		common.LogInfo(c.Request.Context(), fmt.Sprintf("user %d has enough quota %d, trusted and no need to pre-consume", userId, userQuota))
 	}
-	if consumeQuota && preConsumedQuota > 0 {
+	if preConsumedQuota > 0 {
 		err := model.PreConsumeTokenQuota(tokenId, preConsumedQuota)
 		if err != nil {
 			return errorWrapper(err, "pre_consume_token_quota_failed", http.StatusForbidden)
@@ -249,6 +295,13 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
 		}
 		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypeGemini:
+		geminiChatRequest := requestOpenAI2Gemini(textRequest)
+		jsonStr, err := json.Marshal(geminiChatRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
 	case APITypeZhipu:
 		zhipuRequest := requestOpenAI2Zhipu(textRequest)
 		jsonStr, err := json.Marshal(zhipuRequest)
@@ -257,8 +310,41 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 		}
 		requestBody = bytes.NewBuffer(jsonStr)
 	case APITypeAli:
-		aliRequest := requestOpenAI2Ali(textRequest)
-		jsonStr, err := json.Marshal(aliRequest)
+		var jsonStr []byte
+		var err error
+		switch relayMode {
+		case RelayModeEmbeddings:
+			aliEmbeddingRequest := embeddingRequestOpenAI2Ali(textRequest)
+			jsonStr, err = json.Marshal(aliEmbeddingRequest)
+		default:
+			aliRequest := requestOpenAI2Ali(textRequest)
+			jsonStr, err = json.Marshal(aliRequest)
+		}
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypeTencent:
+		apiKey := c.Request.Header.Get("Authorization")
+		apiKey = strings.TrimPrefix(apiKey, "Bearer ")
+		appId, secretId, secretKey, err := parseTencentConfig(apiKey)
+		if err != nil {
+			return errorWrapper(err, "invalid_tencent_config", http.StatusInternalServerError)
+		}
+		tencentRequest := requestOpenAI2Tencent(textRequest)
+		tencentRequest.AppId = appId
+		tencentRequest.SecretId = secretId
+		jsonStr, err := json.Marshal(tencentRequest)
+		if err != nil {
+			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
+		}
+		sign := getTencentSign(*tencentRequest, secretKey)
+		c.Request.Header.Set("Authorization", sign)
+		requestBody = bytes.NewBuffer(jsonStr)
+	case APITypeAIProxyLibrary:
+		aiProxyLibraryRequest := requestOpenAI2AIProxyLibrary(textRequest)
+		aiProxyLibraryRequest.LibraryId = c.GetString("library_id")
+		jsonStr, err := json.Marshal(aiProxyLibraryRequest)
 		if err != nil {
 			return errorWrapper(err, "marshal_text_request_failed", http.StatusInternalServerError)
 		}
@@ -282,6 +368,10 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 				req.Header.Set("api-key", apiKey)
 			} else {
 				req.Header.Set("Authorization", c.Request.Header.Get("Authorization"))
+				if channelType == common.ChannelTypeOpenRouter {
+					req.Header.Set("HTTP-Referer", "https://github.com/songquanpeng/one-api")
+					req.Header.Set("X-Title", "One API")
+				}
 			}
 		case APITypeClaude:
 			req.Header.Set("x-api-key", apiKey)
@@ -298,9 +388,23 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			if textRequest.Stream {
 				req.Header.Set("X-DashScope-SSE", "enable")
 			}
+			if c.GetString("plugin") != "" {
+				req.Header.Set("X-DashScope-Plugin", c.GetString("plugin"))
+			}
+		case APITypeTencent:
+			req.Header.Set("Authorization", apiKey)
+		case APITypePaLM:
+			// do not set Authorization header
+		case APITypeGemini:
+			// do not set Authorization header
+		default:
+			req.Header.Set("Authorization", "Bearer "+apiKey)
 		}
 		req.Header.Set("Content-Type", c.Request.Header.Get("Content-Type"))
 		req.Header.Set("Accept", c.Request.Header.Get("Accept"))
+		if isStream && c.Request.Header.Get("Accept") == "" {
+			req.Header.Set("Accept", "text/event-stream")
+		}
 		//req.Header.Set("Connection", c.Request.Header.Get("Connection"))
 		resp, err = httpClient.Do(req)
 		if err != nil {
@@ -317,53 +421,57 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 		isStream = isStream || strings.HasPrefix(resp.Header.Get("Content-Type"), "text/event-stream")
 
 		if resp.StatusCode != http.StatusOK {
+			if preConsumedQuota != 0 {
+				go func(ctx context.Context) {
+					// return pre-consumed quota
+					err := model.PostConsumeTokenQuota(tokenId, -preConsumedQuota)
+					if err != nil {
+						common.LogError(ctx, "error return pre-consumed quota: "+err.Error())
+					}
+				}(c.Request.Context())
+			}
 			return relayErrorHandler(resp)
 		}
 	}
 
 	var textResponse TextResponse
 	tokenName := c.GetString("token_name")
-	channelId := c.GetInt("channel_id")
 
-	defer func() {
+	defer func(ctx context.Context) {
 		// c.Writer.Flush()
 		go func() {
-			if consumeQuota {
-				quota := 0
-				completionRatio := common.GetCompletionRatio(textRequest.Model)
-				promptTokens = textResponse.Usage.PromptTokens
-				completionTokens = textResponse.Usage.CompletionTokens
-
-				quota = promptTokens + int(float64(completionTokens)*completionRatio)
-				quota = int(float64(quota) * ratio)
-				if ratio != 0 && quota <= 0 {
-					quota = 1
-				}
-				totalTokens := promptTokens + completionTokens
-				if totalTokens == 0 {
-					// in this case, must be some error happened
-					// we cannot just return, because we may have to return the pre-consumed quota
-					quota = 0
-				}
-				quotaDelta := quota - preConsumedQuota
-				err := model.PostConsumeTokenQuota(tokenId, quotaDelta)
-				if err != nil {
-					common.SysError("error consuming token remain quota: " + err.Error())
-				}
-				err = model.CacheUpdateUserQuota(userId)
-				if err != nil {
-					common.SysError("error update user quota cache: " + err.Error())
-				}
-				if quota != 0 {
-					logContent := fmt.Sprintf("模型倍率 %.2f，分组倍率 %.2f", modelRatio, groupRatio)
-					model.RecordConsumeLog(userId, promptTokens, completionTokens, textRequest.Model, tokenName, quota, logContent)
-					model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
-
-					model.UpdateChannelUsedQuota(channelId, quota)
-				}
+			quota := 0
+			completionRatio := common.GetCompletionRatio(textRequest.Model)
+			promptTokens = textResponse.Usage.PromptTokens
+			completionTokens = textResponse.Usage.CompletionTokens
+			quota = int(math.Ceil((float64(promptTokens) + float64(completionTokens)*completionRatio) * ratio))
+			if ratio != 0 && quota <= 0 {
+				quota = 1
 			}
+			totalTokens := promptTokens + completionTokens
+			if totalTokens == 0 {
+				// in this case, must be some error happened
+				// we cannot just return, because we may have to return the pre-consumed quota
+				quota = 0
+			}
+			quotaDelta := quota - preConsumedQuota
+			err := model.PostConsumeTokenQuota(tokenId, quotaDelta)
+			if err != nil {
+				common.LogError(ctx, "error consuming token remain quota: "+err.Error())
+			}
+			err = model.CacheUpdateUserQuota(userId)
+			if err != nil {
+				common.LogError(ctx, "error update user quota cache: "+err.Error())
+			}
+			if quota != 0 {
+				logContent := fmt.Sprintf("模型倍率 %.2f，分组倍率 %.2f", modelRatio, groupRatio)
+				model.RecordConsumeLog(ctx, userId, channelId, promptTokens, completionTokens, textRequest.Model, tokenName, quota, logContent)
+				model.UpdateUserUsedQuotaAndRequestCount(userId, quota)
+				model.UpdateChannelUsedQuota(channelId, quota)
+			}
+
 		}()
-	}()
+	}(c.Request.Context())
 	switch apiType {
 	case APITypeOpenAI:
 		if isStream {
@@ -375,7 +483,7 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			textResponse.Usage.CompletionTokens = countTokenText(responseText, textRequest.Model)
 			return nil
 		} else {
-			err, usage := openaiHandler(c, resp, consumeQuota, promptTokens, textRequest.Model)
+			err, usage := openaiHandler(c, resp, promptTokens, textRequest.Model)
 			if err != nil {
 				return err
 			}
@@ -449,6 +557,25 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			}
 			return nil
 		}
+	case APITypeGemini:
+		if textRequest.Stream {
+			err, responseText := geminiChatStreamHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			textResponse.Usage.PromptTokens = promptTokens
+			textResponse.Usage.CompletionTokens = countTokenText(responseText, textRequest.Model)
+			return nil
+		} else {
+			err, usage := geminiChatHandler(c, resp, promptTokens, textRequest.Model)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
+		}
 	case APITypeZhipu:
 		if isStream {
 			err, usage := zhipuStreamHandler(c, resp)
@@ -484,7 +611,14 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			}
 			return nil
 		} else {
-			err, usage := aliHandler(c, resp)
+			var err *OpenAIErrorWithStatusCode
+			var usage *Usage
+			switch relayMode {
+			case RelayModeEmbeddings:
+				err, usage = aliEmbeddingHandler(c, resp)
+			default:
+				err, usage = aliHandler(c, resp)
+			}
 			if err != nil {
 				return err
 			}
@@ -494,14 +628,29 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			return nil
 		}
 	case APITypeXunfei:
+		auth := c.Request.Header.Get("Authorization")
+		auth = strings.TrimPrefix(auth, "Bearer ")
+		splits := strings.Split(auth, "|")
+		if len(splits) != 3 {
+			return errorWrapper(errors.New("invalid auth"), "invalid_auth", http.StatusBadRequest)
+		}
+		var err *OpenAIErrorWithStatusCode
+		var usage *Usage
 		if isStream {
-			auth := c.Request.Header.Get("Authorization")
-			auth = strings.TrimPrefix(auth, "Bearer ")
-			splits := strings.Split(auth, "|")
-			if len(splits) != 3 {
-				return errorWrapper(errors.New("invalid auth"), "invalid_auth", http.StatusBadRequest)
-			}
-			err, usage := xunfeiStreamHandler(c, textRequest, splits[0], splits[1], splits[2])
+			err, usage = xunfeiStreamHandler(c, textRequest, splits[0], splits[1], splits[2])
+		} else {
+			err, usage = xunfeiHandler(c, textRequest, splits[0], splits[1], splits[2])
+		}
+		if err != nil {
+			return err
+		}
+		if usage != nil {
+			textResponse.Usage = *usage
+		}
+		return nil
+	case APITypeAIProxyLibrary:
+		if isStream {
+			err, usage := aiProxyLibraryStreamHandler(c, resp)
 			if err != nil {
 				return err
 			}
@@ -510,7 +659,33 @@ func relayTextHelper(c *gin.Context, relayMode int) *OpenAIErrorWithStatusCode {
 			}
 			return nil
 		} else {
-			return errorWrapper(errors.New("xunfei api does not support non-stream mode"), "invalid_api_type", http.StatusBadRequest)
+			err, usage := aiProxyLibraryHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
+		}
+	case APITypeTencent:
+		if isStream {
+			err, responseText := tencentStreamHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			textResponse.Usage.PromptTokens = promptTokens
+			textResponse.Usage.CompletionTokens = countTokenText(responseText, textRequest.Model)
+			return nil
+		} else {
+			err, usage := tencentHandler(c, resp)
+			if err != nil {
+				return err
+			}
+			if usage != nil {
+				textResponse.Usage = *usage
+			}
+			return nil
 		}
 	default:
 		return errorWrapper(errors.New("unknown api type"), "unknown_api_type", http.StatusInternalServerError)

controller/relay-utils.go

@@ -1,52 +1,67 @@
 package controller
 
 import (
+	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
-	"github.com/pkoukk/tiktoken-go"
 	"io"
+	"math"
 	"net/http"
 	"one-api/common"
+	"one-api/common/image"
+	"one-api/model"
 	"strconv"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+	"github.com/pkoukk/tiktoken-go"
 )
 
 var stopFinishReason = "stop"
 
+// tokenEncoderMap won't grow after initialization
 var tokenEncoderMap = map[string]*tiktoken.Tiktoken{}
+var defaultTokenEncoder *tiktoken.Tiktoken
 
 func InitTokenEncoders() {
 	common.SysLog("initializing token encoders")
-	fallbackTokenEncoder, err := tiktoken.EncodingForModel("gpt-3.5-turbo")
+	gpt35TokenEncoder, err := tiktoken.EncodingForModel("gpt-3.5-turbo")
 	if err != nil {
-		common.FatalLog(fmt.Sprintf("failed to get fallback token encoder: %s", err.Error()))
+		common.FatalLog(fmt.Sprintf("failed to get gpt-3.5-turbo token encoder: %s", err.Error()))
+	}
+	defaultTokenEncoder = gpt35TokenEncoder
+	gpt4TokenEncoder, err := tiktoken.EncodingForModel("gpt-4")
+	if err != nil {
+		common.FatalLog(fmt.Sprintf("failed to get gpt-4 token encoder: %s", err.Error()))
 	}
 	for model, _ := range common.ModelRatio {
-		tokenEncoder, err := tiktoken.EncodingForModel(model)
-		if err != nil {
-			common.SysError(fmt.Sprintf("using fallback encoder for model %s", model))
-			tokenEncoderMap[model] = fallbackTokenEncoder
-			continue
+		if strings.HasPrefix(model, "gpt-3.5") {
+			tokenEncoderMap[model] = gpt35TokenEncoder
+		} else if strings.HasPrefix(model, "gpt-4") {
+			tokenEncoderMap[model] = gpt4TokenEncoder
+		} else {
+			tokenEncoderMap[model] = nil
 		}
-		tokenEncoderMap[model] = tokenEncoder
 	}
 	common.SysLog("token encoders initialized")
 }
 
 func getTokenEncoder(model string) *tiktoken.Tiktoken {
-	if tokenEncoder, ok := tokenEncoderMap[model]; ok {
+	tokenEncoder, ok := tokenEncoderMap[model]
+	if ok && tokenEncoder != nil {
 		return tokenEncoder
 	}
-	tokenEncoder, err := tiktoken.EncodingForModel(model)
-	if err != nil {
-		common.SysError(fmt.Sprintf("failed to get token encoder for model %s: %s, using encoder for gpt-3.5-turbo", model, err.Error()))
-		tokenEncoder, err = tiktoken.EncodingForModel("gpt-3.5-turbo")
+	if ok {
+		tokenEncoder, err := tiktoken.EncodingForModel(model)
 		if err != nil {
-			common.FatalLog(fmt.Sprintf("failed to get token encoder for model gpt-3.5-turbo: %s", err.Error()))
+			common.SysError(fmt.Sprintf("failed to get token encoder for model %s: %s, using encoder for gpt-3.5-turbo", model, err.Error()))
+			tokenEncoder = defaultTokenEncoder
 		}
+		tokenEncoderMap[model] = tokenEncoder
+		return tokenEncoder
 	}
-	tokenEncoderMap[model] = tokenEncoder
-	return tokenEncoder
+	return defaultTokenEncoder
 }
 
 func getTokenNum(tokenEncoder *tiktoken.Tiktoken, text string) int {
@@ -75,7 +90,33 @@ func countTokenMessages(messages []Message, model string) int {
 	tokenNum := 0
 	for _, message := range messages {
 		tokenNum += tokensPerMessage
-		tokenNum += getTokenNum(tokenEncoder, message.Content)
+		switch v := message.Content.(type) {
+		case string:
+			tokenNum += getTokenNum(tokenEncoder, v)
+		case []any:
+			for _, it := range v {
+				m := it.(map[string]any)
+				switch m["type"] {
+				case "text":
+					tokenNum += getTokenNum(tokenEncoder, m["text"].(string))
+				case "image_url":
+					imageUrl, ok := m["image_url"].(map[string]any)
+					if ok {
+						url := imageUrl["url"].(string)
+						detail := ""
+						if imageUrl["detail"] != nil {
+							detail = imageUrl["detail"].(string)
+						}
+						imageTokens, err := countImageTokens(url, detail)
+						if err != nil {
+							common.SysError("error counting image tokens: " + err.Error())
+						} else {
+							tokenNum += imageTokens
+						}
+					}
+				}
+			}
+		}
 		tokenNum += getTokenNum(tokenEncoder, message.Role)
 		if message.Name != nil {
 			tokenNum += tokensPerName
@@ -86,13 +127,81 @@ func countTokenMessages(messages []Message, model string) int {
 	return tokenNum
 }
 
+const (
+	lowDetailCost         = 85
+	highDetailCostPerTile = 170
+	additionalCost        = 85
+)
+
+// https://platform.openai.com/docs/guides/vision/calculating-costs
+// https://github.com/openai/openai-cookbook/blob/05e3f9be4c7a2ae7ecf029a7c32065b024730ebe/examples/How_to_count_tokens_with_tiktoken.ipynb
+func countImageTokens(url string, detail string) (_ int, err error) {
+	var fetchSize = true
+	var width, height int
+	// Reference: https://platform.openai.com/docs/guides/vision/low-or-high-fidelity-image-understanding
+	// detail == "auto" is undocumented on how it works, it just said the model will use the auto setting which will look at the image input size and decide if it should use the low or high setting.
+	// According to the official guide, "low" disable the high-res model,
+	// and only receive low-res 512px x 512px version of the image, indicating
+	// that image is treated as low-res when size is smaller than 512px x 512px,
+	// then we can assume that image size larger than 512px x 512px is treated
+	// as high-res. Then we have the following logic:
+	// if detail == "" || detail == "auto" {
+	// 	width, height, err = image.GetImageSize(url)
+	// 	if err != nil {
+	// 		return 0, err
+	// 	}
+	// 	fetchSize = false
+	// 	// not sure if this is correct
+	// 	if width > 512 || height > 512 {
+	// 		detail = "high"
+	// 	} else {
+	// 		detail = "low"
+	// 	}
+	// }
+
+	// However, in my test, it seems to be always the same as "high".
+	// The following image, which is 125x50, is still treated as high-res, taken
+	// 255 tokens in the response of non-stream chat completion api.
+	// https://upload.wikimedia.org/wikipedia/commons/1/10/18_Infantry_Division_Messina.jpg
+	if detail == "" || detail == "auto" {
+		// assume by test, not sure if this is correct
+		detail = "high"
+	}
+	switch detail {
+	case "low":
+		return lowDetailCost, nil
+	case "high":
+		if fetchSize {
+			width, height, err = image.GetImageSize(url)
+			if err != nil {
+				return 0, err
+			}
+		}
+		if width > 2048 || height > 2048 { // max(width, height) > 2048
+			ratio := float64(2048) / math.Max(float64(width), float64(height))
+			width = int(float64(width) * ratio)
+			height = int(float64(height) * ratio)
+		}
+		if width > 768 && height > 768 { // min(width, height) > 768
+			ratio := float64(768) / math.Min(float64(width), float64(height))
+			width = int(float64(width) * ratio)
+			height = int(float64(height) * ratio)
+		}
+		numSquares := int(math.Ceil(float64(width)/512) * math.Ceil(float64(height)/512))
+		result := numSquares*highDetailCostPerTile + additionalCost
+		return result, nil
+	default:
+		return 0, errors.New("invalid detail option")
+	}
+}
+
 func countTokenInput(input any, model string) int {
-	switch input.(type) {
+	switch v := input.(type) {
 	case string:
-		return countTokenText(input.(string), model)
+		return countTokenText(v, model)
 	case []string:
 		text := ""
-		for _, s := range input.([]string) {
+		for _, s := range v {
 			text += s
 		}
 		return countTokenText(text, model)
@@ -133,6 +242,19 @@ func shouldDisableChannel(err *OpenAIError, statusCode int) bool {
 	return false
 }
 
+func shouldEnableChannel(err error, openAIErr *OpenAIError) bool {
+	if !common.AutomaticEnableChannelEnabled {
+		return false
+	}
+	if err != nil {
+		return false
+	}
+	if openAIErr != nil {
+		return false
+	}
+	return true
+}
+
 func setEventStreamHeaders(c *gin.Context) {
 	c.Writer.Header().Set("Content-Type", "text/event-stream")
 	c.Writer.Header().Set("Cache-Control", "no-cache")
@@ -146,7 +268,7 @@ func relayErrorHandler(resp *http.Response) (openAIErrorWithStatusCode *OpenAIEr
 		StatusCode: resp.StatusCode,
 		OpenAIError: OpenAIError{
 			Message: fmt.Sprintf("bad response status code %d", resp.StatusCode),
-			Type:    "one_api_error",
+			Type:    "upstream_error",
 			Code:    "bad_response_status_code",
 			Param:   strconv.Itoa(resp.StatusCode),
 		},
@@ -167,3 +289,48 @@ func relayErrorHandler(resp *http.Response) (openAIErrorWithStatusCode *OpenAIEr
 	openAIErrorWithStatusCode.OpenAIError = textResponse.Error
 	return
 }
+
+func getFullRequestURL(baseURL string, requestURL string, channelType int) string {
+	fullRequestURL := fmt.Sprintf("%s%s", baseURL, requestURL)
+
+	if strings.HasPrefix(baseURL, "https://gateway.ai.cloudflare.com") {
+		switch channelType {
+		case common.ChannelTypeOpenAI:
+			fullRequestURL = fmt.Sprintf("%s%s", baseURL, strings.TrimPrefix(requestURL, "/v1"))
+		case common.ChannelTypeAzure:
+			fullRequestURL = fmt.Sprintf("%s%s", baseURL, strings.TrimPrefix(requestURL, "/openai/deployments"))
+		}
+	}
+	return fullRequestURL
+}
+
+func postConsumeQuota(ctx context.Context, tokenId int, quotaDelta int, totalQuota int, userId int, channelId int, modelRatio float64, groupRatio float64, modelName string, tokenName string) {
+	// quotaDelta is remaining quota to be consumed
+	err := model.PostConsumeTokenQuota(tokenId, quotaDelta)
+	if err != nil {
+		common.SysError("error consuming token remain quota: " + err.Error())
+	}
+	err = model.CacheUpdateUserQuota(userId)
+	if err != nil {
+		common.SysError("error update user quota cache: " + err.Error())
+	}
+	// totalQuota is total quota consumed
+	if totalQuota != 0 {
+		logContent := fmt.Sprintf("模型倍率 %.2f，分组倍率 %.2f", modelRatio, groupRatio)
+		model.RecordConsumeLog(ctx, userId, channelId, totalQuota, 0, modelName, tokenName, totalQuota, logContent)
+		model.UpdateUserUsedQuotaAndRequestCount(userId, totalQuota)
+		model.UpdateChannelUsedQuota(channelId, totalQuota)
+	}
+	if totalQuota <= 0 {
+		common.LogError(ctx, fmt.Sprintf("totalQuota consumed is %d, something is wrong", totalQuota))
+	}
+}
+
+func GetAPIVersion(c *gin.Context) string {
+	query := c.Request.URL.Query()
+	apiVersion := query.Get("api-version")
+	if apiVersion == "" {
+		apiVersion = c.GetString("api_version")
+	}
+	return apiVersion
+}

controller/relay-xunfei.go

@@ -81,7 +81,7 @@ func requestOpenAI2Xunfei(request GeneralOpenAIRequest, xunfeiAppId string, doma
 		if message.Role == "system" {
 			messages = append(messages, XunfeiMessage{
 				Role:    "user",
-				Content: message.Content,
+				Content: message.StringContent(),
 			})
 			messages = append(messages, XunfeiMessage{
 				Role:    "assistant",
@@ -90,7 +90,7 @@ func requestOpenAI2Xunfei(request GeneralOpenAIRequest, xunfeiAppId string, doma
 		} else {
 			messages = append(messages, XunfeiMessage{
 				Role:    message.Role,
-				Content: message.Content,
+				Content: message.StringContent(),
 			})
 		}
 	}
@@ -118,6 +118,7 @@ func responseXunfei2OpenAI(response *XunfeiChatResponse) *OpenAITextResponse {
 			Role:    "assistant",
 			Content: response.Payload.Choices.Text[0].Content,
 		},
+		FinishReason: stopFinishReason,
 	}
 	fullTextResponse := OpenAITextResponse{
 		Object:  "chat.completion",
@@ -177,33 +178,85 @@ func buildXunfeiAuthUrl(hostUrl string, apiKey, apiSecret string) string {
 }
 
 func xunfeiStreamHandler(c *gin.Context, textRequest GeneralOpenAIRequest, appId string, apiSecret string, apiKey string) (*OpenAIErrorWithStatusCode, *Usage) {
+	domain, authUrl := getXunfeiAuthUrl(c, apiKey, apiSecret)
+	dataChan, stopChan, err := xunfeiMakeRequest(textRequest, domain, authUrl, appId)
+	if err != nil {
+		return errorWrapper(err, "make xunfei request err", http.StatusInternalServerError), nil
+	}
+	setEventStreamHeaders(c)
 	var usage Usage
-	query := c.Request.URL.Query()
-	apiVersion := query.Get("api-version")
-	if apiVersion == "" {
-		apiVersion = c.GetString("api_version")
+	c.Stream(func(w io.Writer) bool {
+		select {
+		case xunfeiResponse := <-dataChan:
+			usage.PromptTokens += xunfeiResponse.Payload.Usage.Text.PromptTokens
+			usage.CompletionTokens += xunfeiResponse.Payload.Usage.Text.CompletionTokens
+			usage.TotalTokens += xunfeiResponse.Payload.Usage.Text.TotalTokens
+			response := streamResponseXunfei2OpenAI(&xunfeiResponse)
+			jsonResponse, err := json.Marshal(response)
+			if err != nil {
+				common.SysError("error marshalling stream response: " + err.Error())
+				return true
+			}
+			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
+			return true
+		case <-stopChan:
+			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
+			return false
+		}
+	})
+	return nil, &usage
+}
+
+func xunfeiHandler(c *gin.Context, textRequest GeneralOpenAIRequest, appId string, apiSecret string, apiKey string) (*OpenAIErrorWithStatusCode, *Usage) {
+	domain, authUrl := getXunfeiAuthUrl(c, apiKey, apiSecret)
+	dataChan, stopChan, err := xunfeiMakeRequest(textRequest, domain, authUrl, appId)
+	if err != nil {
+		return errorWrapper(err, "make xunfei request err", http.StatusInternalServerError), nil
 	}
-	if apiVersion == "" {
-		apiVersion = "v1.1"
-		common.SysLog("api_version not found, use default: " + apiVersion)
+	var usage Usage
+	var content string
+	var xunfeiResponse XunfeiChatResponse
+	stop := false
+	for !stop {
+		select {
+		case xunfeiResponse = <-dataChan:
+			if len(xunfeiResponse.Payload.Choices.Text) == 0 {
+				continue
+			}
+			content += xunfeiResponse.Payload.Choices.Text[0].Content
+			usage.PromptTokens += xunfeiResponse.Payload.Usage.Text.PromptTokens
+			usage.CompletionTokens += xunfeiResponse.Payload.Usage.Text.CompletionTokens
+			usage.TotalTokens += xunfeiResponse.Payload.Usage.Text.TotalTokens
+		case stop = <-stopChan:
+		}
 	}
-	domain := "general"
-	if apiVersion == "v2.1" {
-		domain = "generalv2"
+
+	xunfeiResponse.Payload.Choices.Text[0].Content = content
+
+	response := responseXunfei2OpenAI(&xunfeiResponse)
+	jsonResponse, err := json.Marshal(response)
+	if err != nil {
+		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
 	}
-	hostUrl := fmt.Sprintf("wss://spark-api.xf-yun.com/%s/chat", apiVersion)
+	c.Writer.Header().Set("Content-Type", "application/json")
+	_, _ = c.Writer.Write(jsonResponse)
+	return nil, &usage
+}
+
+func xunfeiMakeRequest(textRequest GeneralOpenAIRequest, domain, authUrl, appId string) (chan XunfeiChatResponse, chan bool, error) {
 	d := websocket.Dialer{
 		HandshakeTimeout: 5 * time.Second,
 	}
-	conn, resp, err := d.Dial(buildXunfeiAuthUrl(hostUrl, apiKey, apiSecret), nil)
+	conn, resp, err := d.Dial(authUrl, nil)
 	if err != nil || resp.StatusCode != 101 {
-		return errorWrapper(err, "dial_failed", http.StatusInternalServerError), nil
+		return nil, nil, err
 	}
 	data := requestOpenAI2Xunfei(textRequest, appId, domain)
 	err = conn.WriteJSON(data)
 	if err != nil {
-		return errorWrapper(err, "write_json_failed", http.StatusInternalServerError), nil
+		return nil, nil, err
 	}
+
 	dataChan := make(chan XunfeiChatResponse)
 	stopChan := make(chan bool)
 	go func() {
@@ -230,61 +283,24 @@ func xunfeiStreamHandler(c *gin.Context, textRequest GeneralOpenAIRequest, appId
 		}
 		stopChan <- true
 	}()
-	setEventStreamHeaders(c)
-	c.Stream(func(w io.Writer) bool {
-		select {
-		case xunfeiResponse := <-dataChan:
-			usage.PromptTokens += xunfeiResponse.Payload.Usage.Text.PromptTokens
-			usage.CompletionTokens += xunfeiResponse.Payload.Usage.Text.CompletionTokens
-			usage.TotalTokens += xunfeiResponse.Payload.Usage.Text.TotalTokens
-			response := streamResponseXunfei2OpenAI(&xunfeiResponse)
-			jsonResponse, err := json.Marshal(response)
-			if err != nil {
-				common.SysError("error marshalling stream response: " + err.Error())
-				return true
-			}
-			c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonResponse)})
-			return true
-		case <-stopChan:
-			c.Render(-1, common.CustomEvent{Data: "data: [DONE]"})
-			return false
-		}
-	})
-	return nil, &usage
+
+	return dataChan, stopChan, nil
 }
 
-func xunfeiHandler(c *gin.Context, resp *http.Response) (*OpenAIErrorWithStatusCode, *Usage) {
-	var xunfeiResponse XunfeiChatResponse
-	responseBody, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return errorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+func getXunfeiAuthUrl(c *gin.Context, apiKey string, apiSecret string) (string, string) {
+	query := c.Request.URL.Query()
+	apiVersion := query.Get("api-version")
+	if apiVersion == "" {
+		apiVersion = c.GetString("api_version")
 	}
-	err = resp.Body.Close()
-	if err != nil {
-		return errorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	if apiVersion == "" {
+		apiVersion = "v1.1"
+		common.SysLog("api_version not found, use default: " + apiVersion)
 	}
-	err = json.Unmarshal(responseBody, &xunfeiResponse)
-	if err != nil {
-		return errorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	domain := "general"
+	if apiVersion != "v1.1" {
+		domain += strings.Split(apiVersion, ".")[0]
 	}
-	if xunfeiResponse.Header.Code != 0 {
-		return &OpenAIErrorWithStatusCode{
-			OpenAIError: OpenAIError{
-				Message: xunfeiResponse.Header.Message,
-				Type:    "xunfei_error",
-				Param:   "",
-				Code:    xunfeiResponse.Header.Code,
-			},
-			StatusCode: resp.StatusCode,
-		}, nil
-	}
-	fullTextResponse := responseXunfei2OpenAI(&xunfeiResponse)
-	jsonResponse, err := json.Marshal(fullTextResponse)
-	if err != nil {
-		return errorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
-	}
-	c.Writer.Header().Set("Content-Type", "application/json")
-	c.Writer.WriteHeader(resp.StatusCode)
-	_, err = c.Writer.Write(jsonResponse)
-	return nil, &fullTextResponse.Usage
+	authUrl := buildXunfeiAuthUrl(fmt.Sprintf("wss://spark-api.xf-yun.com/%s/chat", apiVersion), apiKey, apiSecret)
+	return domain, authUrl
 }

controller/relay-zhipu.go

@@ -114,7 +114,7 @@ func requestOpenAI2Zhipu(request GeneralOpenAIRequest) *ZhipuRequest {
 		if message.Role == "system" {
 			messages = append(messages, ZhipuMessage{
 				Role:    "system",
-				Content: message.Content,
+				Content: message.StringContent(),
 			})
 			messages = append(messages, ZhipuMessage{
 				Role:    "user",
@@ -123,7 +123,7 @@ func requestOpenAI2Zhipu(request GeneralOpenAIRequest) *ZhipuRequest {
 		} else {
 			messages = append(messages, ZhipuMessage{
 				Role:    message.Role,
-				Content: message.Content,
+				Content: message.StringContent(),
 			})
 		}
 	}

controller/relay.go

@@ -12,10 +12,49 @@ import (
 
 type Message struct {
 	Role    string  `json:"role"`
-	Content string  `json:"content"`
+	Content any     `json:"content"`
 	Name    *string `json:"name,omitempty"`
 }
 
+type ImageURL struct {
+	Url    string `json:"url,omitempty"`
+	Detail string `json:"detail,omitempty"`
+}
+
+type TextContent struct {
+	Type string `json:"type,omitempty"`
+	Text string `json:"text,omitempty"`
+}
+
+type ImageContent struct {
+	Type     string    `json:"type,omitempty"`
+	ImageURL *ImageURL `json:"image_url,omitempty"`
+}
+
+func (m Message) StringContent() string {
+	content, ok := m.Content.(string)
+	if ok {
+		return content
+	}
+	contentList, ok := m.Content.([]any)
+	if ok {
+		var contentStr string
+		for _, contentItem := range contentList {
+			contentMap, ok := contentItem.(map[string]any)
+			if !ok {
+				continue
+			}
+			if contentMap["type"] == "text" {
+				if subStr, ok := contentMap["text"].(string); ok {
+					contentStr += subStr
+				}
+			}
+		}
+		return contentStr
+	}
+	return ""
+}
+
 const (
 	RelayModeUnknown = iota
 	RelayModeChatCompletions
@@ -24,23 +63,56 @@ const (
 	RelayModeModerations
 	RelayModeImagesGenerations
 	RelayModeEdits
+	RelayModeAudioSpeech
+	RelayModeAudioTranscription
+	RelayModeAudioTranslation
 )
 
 // https://platform.openai.com/docs/api-reference/chat
 
+type ResponseFormat struct {
+	Type string `json:"type,omitempty"`
+}
+
 type GeneralOpenAIRequest struct {
-	Model       string    `json:"model,omitempty"`
-	Messages    []Message `json:"messages,omitempty"`
-	Prompt      any       `json:"prompt,omitempty"`
-	Stream      bool      `json:"stream,omitempty"`
-	MaxTokens   int       `json:"max_tokens,omitempty"`
-	Temperature float64   `json:"temperature,omitempty"`
-	TopP        float64   `json:"top_p,omitempty"`
-	N           int       `json:"n,omitempty"`
-	Input       any       `json:"input,omitempty"`
-	Instruction string    `json:"instruction,omitempty"`
-	Size        string    `json:"size,omitempty"`
-	Functions   any       `json:"functions,omitempty"`
+	Model            string          `json:"model,omitempty"`
+	Messages         []Message       `json:"messages,omitempty"`
+	Prompt           any             `json:"prompt,omitempty"`
+	Stream           bool            `json:"stream,omitempty"`
+	MaxTokens        int             `json:"max_tokens,omitempty"`
+	Temperature      float64         `json:"temperature,omitempty"`
+	TopP             float64         `json:"top_p,omitempty"`
+	N                int             `json:"n,omitempty"`
+	Input            any             `json:"input,omitempty"`
+	Instruction      string          `json:"instruction,omitempty"`
+	Size             string          `json:"size,omitempty"`
+	Functions        any             `json:"functions,omitempty"`
+	FrequencyPenalty float64         `json:"frequency_penalty,omitempty"`
+	PresencePenalty  float64         `json:"presence_penalty,omitempty"`
+	ResponseFormat   *ResponseFormat `json:"response_format,omitempty"`
+	Seed             float64         `json:"seed,omitempty"`
+	Tools            any             `json:"tools,omitempty"`
+	ToolChoice       any             `json:"tool_choice,omitempty"`
+	User             string          `json:"user,omitempty"`
+}
+
+func (r GeneralOpenAIRequest) ParseInput() []string {
+	if r.Input == nil {
+		return nil
+	}
+	var input []string
+	switch r.Input.(type) {
+	case string:
+		input = []string{r.Input.(string)}
+	case []any:
+		input = make([]string, 0, len(r.Input.([]any)))
+		for _, item := range r.Input.([]any) {
+			if str, ok := item.(string); ok {
+				input = append(input, str)
+			}
+		}
+	}
+	return input
 }
 
 type ChatRequest struct {
@@ -57,10 +129,49 @@ type TextRequest struct {
 	//Stream   bool      `json:"stream"`
 }
 
+// ImageRequest docs: https://platform.openai.com/docs/api-reference/images/create
 type ImageRequest struct {
-	Prompt string `json:"prompt"`
-	N      int    `json:"n"`
-	Size   string `json:"size"`
+	Model          string `json:"model"`
+	Prompt         string `json:"prompt" binding:"required"`
+	N              int    `json:"n,omitempty"`
+	Size           string `json:"size,omitempty"`
+	Quality        string `json:"quality,omitempty"`
+	ResponseFormat string `json:"response_format,omitempty"`
+	Style          string `json:"style,omitempty"`
+	User           string `json:"user,omitempty"`
+}
+
+type WhisperJSONResponse struct {
+	Text string `json:"text,omitempty"`
+}
+
+type WhisperVerboseJSONResponse struct {
+	Task     string    `json:"task,omitempty"`
+	Language string    `json:"language,omitempty"`
+	Duration float64   `json:"duration,omitempty"`
+	Text     string    `json:"text,omitempty"`
+	Segments []Segment `json:"segments,omitempty"`
+}
+
+type Segment struct {
+	Id               int     `json:"id"`
+	Seek             int     `json:"seek"`
+	Start            float64 `json:"start"`
+	End              float64 `json:"end"`
+	Text             string  `json:"text"`
+	Tokens           []int   `json:"tokens"`
+	Temperature      float64 `json:"temperature"`
+	AvgLogprob       float64 `json:"avg_logprob"`
+	CompressionRatio float64 `json:"compression_ratio"`
+	NoSpeechProb     float64 `json:"no_speech_prob"`
+}
+
+type TextToSpeechRequest struct {
+	Model          string  `json:"model" binding:"required"`
+	Input          string  `json:"input" binding:"required"`
+	Voice          string  `json:"voice" binding:"required"`
+	Speed          float64 `json:"speed"`
+	ResponseFormat string  `json:"response_format"`
 }
 
 type Usage struct {
@@ -159,15 +270,28 @@ func Relay(c *gin.Context) {
 		relayMode = RelayModeImagesGenerations
 	} else if strings.HasPrefix(c.Request.URL.Path, "/v1/edits") {
 		relayMode = RelayModeEdits
+	} else if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/speech") {
+		relayMode = RelayModeAudioSpeech
+	} else if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/transcriptions") {
+		relayMode = RelayModeAudioTranscription
+	} else if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/translations") {
+		relayMode = RelayModeAudioTranslation
 	}
 	var err *OpenAIErrorWithStatusCode
 	switch relayMode {
 	case RelayModeImagesGenerations:
 		err = relayImageHelper(c, relayMode)
+	case RelayModeAudioSpeech:
+		fallthrough
+	case RelayModeAudioTranslation:
+		fallthrough
+	case RelayModeAudioTranscription:
+		err = relayAudioHelper(c, relayMode)
 	default:
 		err = relayTextHelper(c, relayMode)
 	}
 	if err != nil {
+		requestId := c.GetString(common.RequestIdKey)
 		retryTimesStr := c.Query("retry")
 		retryTimes, _ := strconv.Atoi(retryTimesStr)
 		if retryTimesStr == "" {
@@ -179,12 +303,13 @@ func Relay(c *gin.Context) {
 			if err.StatusCode == http.StatusTooManyRequests {
 				err.OpenAIError.Message = "当前分组上游负载已饱和，请稍后再试"
 			}
+			err.OpenAIError.Message = common.MessageWithRequestId(err.OpenAIError.Message, requestId)
 			c.JSON(err.StatusCode, gin.H{
 				"error": err.OpenAIError,
 			})
 		}
 		channelId := c.GetInt("channel_id")
-		common.SysError(fmt.Sprintf("relay error (channel #%d): %s", channelId, err.Message))
+		common.LogError(c.Request.Context(), fmt.Sprintf("relay error (channel #%d): %s", channelId, err.Message))
 		// https://platform.openai.com/docs/guides/error-codes/api-errors
 		if shouldDisableChannel(&err.OpenAIError, err.StatusCode) {
 			channelId := c.GetInt("channel_id")

docker-compose.yml

@@ -9,21 +9,21 @@ services:
     ports:
       - "3000:3000"
     volumes:
-      - ./data:/data
+      - ./data/oneapi:/data
       - ./logs:/app/logs
     environment:
-      - SQL_DSN=root:123456@tcp(host.docker.internal:3306)/one-api  # 修改此行，或注释掉以使用 SQLite 作为数据库
+      - SQL_DSN=oneapi:123456@tcp(db:3306)/one-api  # 修改此行，或注释掉以使用 SQLite 作为数据库
       - REDIS_CONN_STRING=redis://redis
       - SESSION_SECRET=random_string  # 修改为随机字符串
       - TZ=Asia/Shanghai
 #      - NODE_TYPE=slave  # 多机部署时从节点取消注释该行
 #      - SYNC_FREQUENCY=60  # 需要定期从数据库加载数据时取消注释该行
 #      - FRONTEND_BASE_URL=https://openai.justsong.cn  # 多机部署时从节点取消注释该行
-
     depends_on:
       - redis
+      - db
     healthcheck:
-      test: [ "CMD-SHELL", "curl -s http://localhost:3000/api/status | grep -o '\"success\":\\s*true' | awk '{print $2}' | grep 'true'" ]
+      test: [ "CMD-SHELL", "wget -q -O - http://localhost:3000/api/status | grep -o '\"success\":\\s*true' | awk -F: '{print $2}'" ]
       interval: 30s
       timeout: 10s
       retries: 3
@@ -32,3 +32,18 @@ services:
     image: redis:latest
     container_name: redis
     restart: always
+
+  db:
+    image: mysql:8.2.0
+    restart: always
+    container_name: mysql
+    volumes:
+      - ./data/mysql:/var/lib/mysql  # 挂载目录，持久化存储
+    ports:
+      - '3306:3306'
+    environment:
+      TZ: Asia/Shanghai   # 设置时区
+      MYSQL_ROOT_PASSWORD: 'OneAPI@justsong' # 设置 root 用户的密码
+      MYSQL_USER: oneapi   # 创建专用用户
+      MYSQL_PASSWORD: '123456'    # 设置专用用户密码
+      MYSQL_DATABASE: one-api   # 自动创建数据库

go.mod

@@ -15,8 +15,11 @@ require (
 	github.com/google/uuid v1.3.0
 	github.com/gorilla/websocket v1.5.0
 	github.com/pkoukk/tiktoken-go v0.1.5
-	golang.org/x/crypto v0.9.0
+	github.com/stretchr/testify v1.8.3
+	golang.org/x/crypto v0.14.0
+	golang.org/x/image v0.14.0
 	gorm.io/driver/mysql v1.4.3
+	gorm.io/driver/postgres v1.5.2
 	gorm.io/driver/sqlite v1.4.3
 	gorm.io/gorm v1.25.0
 )
@@ -25,6 +28,7 @@ require (
 	github.com/bytedance/sonic v1.9.1 // indirect
 	github.com/cespare/xxhash/v2 v2.1.2 // indirect
 	github.com/chenzhuoyu/base64x v0.0.0-20221115062448-fe3a3abad311 // indirect
+	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/dgryski/go-rendezvous v0.0.0-20200823014737-9f7001d12a5f // indirect
 	github.com/dlclark/regexp2 v1.10.0 // indirect
 	github.com/gabriel-vasile/mimetype v1.4.2 // indirect
@@ -49,13 +53,13 @@ require (
 	github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect
 	github.com/modern-go/reflect2 v1.0.2 // indirect
 	github.com/pelletier/go-toml/v2 v2.0.8 // indirect
+	github.com/pmezard/go-difflib v1.0.0 // indirect
 	github.com/twitchyliquid64/golang-asm v0.15.1 // indirect
 	github.com/ugorji/go/codec v1.2.11 // indirect
 	golang.org/x/arch v0.3.0 // indirect
-	golang.org/x/net v0.10.0 // indirect
-	golang.org/x/sys v0.8.0 // indirect
-	golang.org/x/text v0.9.0 // indirect
+	golang.org/x/net v0.17.0 // indirect
+	golang.org/x/sys v0.13.0 // indirect
+	golang.org/x/text v0.14.0 // indirect
 	google.golang.org/protobuf v1.30.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
-	gorm.io/driver/postgres v1.5.2 // indirect
 )

go.sum

@@ -150,11 +150,13 @@ golang.org/x/arch v0.0.0-20210923205945-b76863e36670/go.mod h1:5om86z9Hs0C8fWVUu
 golang.org/x/arch v0.3.0 h1:02VY4/ZcO/gBOH6PUaoiptASxtXU10jazRCP865E97k=
 golang.org/x/arch v0.3.0/go.mod h1:5om86z9Hs0C8fWVUuoMHwpExlXzs5Tkyp9hOrfG7pp8=
 golang.org/x/crypto v0.0.0-20210711020723-a769d52b0f97/go.mod h1:GvvjBRRGRdwPK5ydBHafDWAxML/pGHZbMvKqRZ5+Abc=
-golang.org/x/crypto v0.9.0 h1:LF6fAI+IutBocDJ2OT0Q1g8plpYljMZ4+lty+dsqw3g=
-golang.org/x/crypto v0.9.0/go.mod h1:yrmDGqONDYtNj3tH8X9dzUun2m2lzPa9ngI6/RUPGR0=
+golang.org/x/crypto v0.14.0 h1:wBqGXzWJW6m1XrIKlAH0Hs1JJ7+9KBwnIO8v66Q9cHc=
+golang.org/x/crypto v0.14.0/go.mod h1:MVFd36DqK4CsrnJYDkBA3VC4m2GkXAM0PvzMCn4JQf4=
+golang.org/x/image v0.14.0 h1:tNgSxAFe3jC4uYqvZdTr84SZoM1KfwdC9SKIFrLjFn4=
+golang.org/x/image v0.14.0/go.mod h1:HUYqC05R2ZcZ3ejNQsIHQDQiwWM4JBqmm6MKANTp4LE=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
-golang.org/x/net v0.10.0 h1:X2//UzNDwYmtCLn7To6G58Wr6f5ahEAQgKNzv9Y951M=
-golang.org/x/net v0.10.0/go.mod h1:0qNGK6F8kojg2nk9dLZ2mShWaEBan6FAoqfSigmmuDg=
+golang.org/x/net v0.17.0 h1:pVaXccu2ozPjCXewfr1S7xza/zcXTity9cCdXQYSjIM=
+golang.org/x/net v0.17.0/go.mod h1:NxSsAGuq816PNPmqtQdLE42eU2Fs7NoRIZrHJAlaCOE=
 golang.org/x/sys v0.0.0-20200116001909-b77594299b42/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210615035016-665e8c7367d1/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
@@ -162,14 +164,14 @@ golang.org/x/sys v0.0.0-20210630005230-0f9fa26af87c/go.mod h1:oPkhp1MJrh7nUepCBc
 golang.org/x/sys v0.0.0-20210806184541-e5e7981a1069/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.0.0-20220704084225-05e143d24a9e/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.8.0 h1:EBmGv8NaZBZTWvrbjNoL6HVt+IVy3QDQpJs7VRIw3tU=
-golang.org/x/sys v0.8.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
+golang.org/x/sys v0.13.0 h1:Af8nKPmuFypiUBjVoU9V20FiaFXOcuZI21p0ycVYYGE=
+golang.org/x/sys v0.13.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
 golang.org/x/text v0.3.6/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
-golang.org/x/text v0.9.0 h1:2sjJmO8cDvYveuX97RDLsxlyUxLl+GHoLxBiRdHllBE=
-golang.org/x/text v0.9.0/go.mod h1:e1OnstbJyHTd6l/uOt8jFFHp6TRDWZR/bV3emEE/zU8=
+golang.org/x/text v0.14.0 h1:ScX5w1eTa3QqT8oi6+ziP7dTV1S2+ALU0bI+0zXKWiQ=
+golang.org/x/text v0.14.0/go.mod h1:18ZOQIKpY8NJVqYksKHtTdi31H5itFRjB5/qKTNYzSU=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543 h1:E7g+9GITq07hpfrRu66IVDexMakfv52eLZ2CXBWiKr4=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
@@ -198,7 +200,6 @@ gorm.io/driver/postgres v1.5.2/go.mod h1:fmpX0m2I1PKuR7mKZiEluwrP3hbs+ps7JIGMUBp
 gorm.io/driver/sqlite v1.4.3 h1:HBBcZSDnWi5BW3B3rwvVTc510KGkBkexlOg0QrmLUuU=
 gorm.io/driver/sqlite v1.4.3/go.mod h1:0Aq3iPO+v9ZKbcdiz8gLWRw5VOPcBOPUQJFLq5e2ecI=
 gorm.io/gorm v1.23.8/go.mod h1:l2lP/RyAtc1ynaTjFksBde/O8v9oOGIApu2/xRitmZk=
-gorm.io/gorm v1.24.0 h1:j/CoiSm6xpRpmzbFJsQHYj+I8bGYWLXVHeYEyyKlF74=
 gorm.io/gorm v1.24.0/go.mod h1:DVrVomtaYTbqs7gB/x2uVvqnXzv0nqjB396B8cG4dBA=
 gorm.io/gorm v1.25.0 h1:+KtYtb2roDz14EQe4bla8CbQlmb9dN3VejSai3lprfU=
 gorm.io/gorm v1.25.0/go.mod h1:L4uxeKpfBml98NYqVqwAdmV1a2nBtAec/cf3fpucW/k=

i18n/en.json

@@ -119,6 +119,7 @@
   " 年 ": " y ",
   "未测试": "Not tested",
   "通道 ${name} 测试成功，耗时 ${time.toFixed(2)} 秒。": "Channel ${name} test succeeded, time consumed ${time.toFixed(2)} s.",
+  "已成功开始测试所有通道，请刷新页面查看结果。": "All channels have been successfully tested, please refresh the page to view the results.",
   "已成功开始测试所有已启用通道，请刷新页面查看结果。": "All enabled channels have been successfully tested, please refresh the page to view the results.",
   "通道 ${name} 余额更新成功！": "Channel ${name} balance updated successfully!",
   "已更新完毕所有已启用通道余额！": "The balance of all enabled channels has been updated!",
@@ -139,6 +140,7 @@
   "启用": "Enable",
   "编辑": "Edit",
   "添加新的渠道": "Add a new channel",
+  "测试所有通道": "Test all channels",
   "测试所有已启用通道": "Test all enabled channels",
   "更新所有已启用通道余额": "Update the balance of all enabled channels",
   "刷新": "Refresh",
@@ -523,5 +525,6 @@
   "按照如下格式输入：": "Enter in the following format:",
   "模型版本": "Model version",
   "请输入星火大模型版本，注意是接口地址中的版本号，例如：v2.1": "Please enter the version of the Starfire model, note that it is the version number in the interface address, for example: v2.1",
-  "点击查看": "click to view"
+  "点击查看": "click to view",
+  "请确保已在 Azure 上创建了 gpt-35-turbo 模型，并且 apiVersion 已正确填写！": "Please make sure that the gpt-35-turbo model has been created on Azure, and the apiVersion has been filled in correctly!"
 }

main.go

@@ -2,6 +2,7 @@ package main
 
 import (
 	"embed"
+	"fmt"
 	"github.com/gin-contrib/sessions"
 	"github.com/gin-contrib/sessions/cookie"
 	"github.com/gin-gonic/gin"
@@ -21,7 +22,7 @@ var buildFS embed.FS
 var indexPage []byte
 
 func main() {
-	common.SetupGinLog()
+	common.SetupLogger()
 	common.SysLog("One API " + common.Version + " started")
 	if os.Getenv("GIN_MODE") != "debug" {
 		gin.SetMode(gin.ReleaseMode)
@@ -50,18 +51,17 @@ func main() {
 	// Initialize options
 	model.InitOptionMap()
 	if common.RedisEnabled {
+		// for compatibility with old versions
+		common.MemoryCacheEnabled = true
+	}
+	if common.MemoryCacheEnabled {
+		common.SysLog("memory cache enabled")
+		common.SysError(fmt.Sprintf("sync frequency: %d seconds", common.SyncFrequency))
 		model.InitChannelCache()
 	}
-	if os.Getenv("SYNC_FREQUENCY") != "" {
-		frequency, err := strconv.Atoi(os.Getenv("SYNC_FREQUENCY"))
-		if err != nil {
-			common.FatalLog("failed to parse SYNC_FREQUENCY: " + err.Error())
-		}
-		common.SyncFrequency = frequency
-		go model.SyncOptions(frequency)
-		if common.RedisEnabled {
-			go model.SyncChannelCache(frequency)
-		}
+	if common.MemoryCacheEnabled {
+		go model.SyncOptions(common.SyncFrequency)
+		go model.SyncChannelCache(common.SyncFrequency)
 	}
 	if os.Getenv("CHANNEL_UPDATE_FREQUENCY") != "" {
 		frequency, err := strconv.Atoi(os.Getenv("CHANNEL_UPDATE_FREQUENCY"))
@@ -77,14 +77,20 @@ func main() {
 		}
 		go controller.AutomaticallyTestChannels(frequency)
 	}
+	if os.Getenv("BATCH_UPDATE_ENABLED") == "true" {
+		common.BatchUpdateEnabled = true
+		common.SysLog("batch update enabled with interval " + strconv.Itoa(common.BatchUpdateInterval) + "s")
+		model.InitBatchUpdater()
+	}
 	controller.InitTokenEncoders()
 
 	// Initialize HTTP server
-	server := gin.Default()
+	server := gin.New()
+	server.Use(gin.Recovery())
 	// This will cause SSE not to work!!!
 	//server.Use(gzip.Gzip(gzip.DefaultCompression))
-	server.Use(middleware.CORS())
-
+	server.Use(middleware.RequestId())
+	middleware.SetUpLogger(server)
 	// Initialize session store
 	store := cookie.NewStore([]byte(common.SessionSecret))
 	server.Use(sessions.Sessions("session", store))

middleware/auth.go

@@ -91,45 +91,26 @@ func TokenAuth() func(c *gin.Context) {
 		key = parts[0]
 		token, err := model.ValidateUserToken(key)
 		if err != nil {
-			c.JSON(http.StatusUnauthorized, gin.H{
-				"error": gin.H{
-					"message": err.Error(),
-					"type":    "one_api_error",
-				},
-			})
-			c.Abort()
+			abortWithMessage(c, http.StatusUnauthorized, err.Error())
 			return
 		}
-		if !model.CacheIsUserEnabled(token.UserId) {
-			c.JSON(http.StatusForbidden, gin.H{
-				"error": gin.H{
-					"message": "用户已被封禁",
-					"type":    "one_api_error",
-				},
-			})
-			c.Abort()
+		userEnabled, err := model.CacheIsUserEnabled(token.UserId)
+		if err != nil {
+			abortWithMessage(c, http.StatusInternalServerError, err.Error())
+			return
+		}
+		if !userEnabled {
+			abortWithMessage(c, http.StatusForbidden, "用户已被封禁")
 			return
 		}
 		c.Set("id", token.UserId)
 		c.Set("token_id", token.Id)
 		c.Set("token_name", token.Name)
-		requestURL := c.Request.URL.String()
-		consumeQuota := true
-		if strings.HasPrefix(requestURL, "/v1/models") {
-			consumeQuota = false
-		}
-		c.Set("consume_quota", consumeQuota)
 		if len(parts) > 1 {
 			if model.IsAdmin(token.UserId) {
 				c.Set("channelId", parts[1])
 			} else {
-				c.JSON(http.StatusForbidden, gin.H{
-					"error": gin.H{
-						"message": "普通用户不支持指定渠道",
-						"type":    "one_api_error",
-					},
-				})
-				c.Abort()
+				abortWithMessage(c, http.StatusForbidden, "普通用户不支持指定渠道")
 				return
 			}
 		}

middleware/distributor.go

@@ -25,34 +25,16 @@ func Distribute() func(c *gin.Context) {
 		if ok {
 			id, err := strconv.Atoi(channelId.(string))
 			if err != nil {
-				c.JSON(http.StatusBadRequest, gin.H{
-					"error": gin.H{
-						"message": "无效的渠道 ID",
-						"type":    "one_api_error",
-					},
-				})
-				c.Abort()
+				abortWithMessage(c, http.StatusBadRequest, "无效的渠道 Id")
 				return
 			}
 			channel, err = model.GetChannelById(id, true)
 			if err != nil {
-				c.JSON(http.StatusBadRequest, gin.H{
-					"error": gin.H{
-						"message": "无效的渠道 ID",
-						"type":    "one_api_error",
-					},
-				})
-				c.Abort()
+				abortWithMessage(c, http.StatusBadRequest, "无效的渠道 Id")
 				return
 			}
 			if channel.Status != common.ChannelStatusEnabled {
-				c.JSON(http.StatusForbidden, gin.H{
-					"error": gin.H{
-						"message": "该渠道已被禁用",
-						"type":    "one_api_error",
-					},
-				})
-				c.Abort()
+				abortWithMessage(c, http.StatusForbidden, "该渠道已被禁用")
 				return
 			}
 		} else {
@@ -60,13 +42,7 @@ func Distribute() func(c *gin.Context) {
 			var modelRequest ModelRequest
 			err := common.UnmarshalBodyReusable(c, &modelRequest)
 			if err != nil {
-				c.JSON(http.StatusBadRequest, gin.H{
-					"error": gin.H{
-						"message": "无效的请求",
-						"type":    "one_api_error",
-					},
-				})
-				c.Abort()
+				abortWithMessage(c, http.StatusBadRequest, "无效的请求")
 				return
 			}
 			if strings.HasPrefix(c.Request.URL.Path, "/v1/moderations") {
@@ -81,7 +57,12 @@ func Distribute() func(c *gin.Context) {
 			}
 			if strings.HasPrefix(c.Request.URL.Path, "/v1/images/generations") {
 				if modelRequest.Model == "" {
-					modelRequest.Model = "dall-e"
+					modelRequest.Model = "dall-e-2"
+				}
+			}
+			if strings.HasPrefix(c.Request.URL.Path, "/v1/audio/transcriptions") || strings.HasPrefix(c.Request.URL.Path, "/v1/audio/translations") {
+				if modelRequest.Model == "" {
+					modelRequest.Model = "whisper-1"
 				}
 			}
 			channel, err = model.CacheGetRandomSatisfiedChannel(userGroup, modelRequest.Model)
@@ -91,24 +72,27 @@ func Distribute() func(c *gin.Context) {
 					common.SysError(fmt.Sprintf("渠道不存在：%d", channel.Id))
 					message = "数据库一致性已被破坏，请联系管理员"
 				}
-				c.JSON(http.StatusServiceUnavailable, gin.H{
-					"error": gin.H{
-						"message": message,
-						"type":    "one_api_error",
-					},
-				})
-				c.Abort()
+				abortWithMessage(c, http.StatusServiceUnavailable, message)
 				return
 			}
 		}
 		c.Set("channel", channel.Type)
 		c.Set("channel_id", channel.Id)
 		c.Set("channel_name", channel.Name)
-		c.Set("model_mapping", channel.ModelMapping)
+		c.Set("model_mapping", channel.GetModelMapping())
 		c.Request.Header.Set("Authorization", fmt.Sprintf("Bearer %s", channel.Key))
-		c.Set("base_url", channel.BaseURL)
-		if channel.Type == common.ChannelTypeAzure || channel.Type == common.ChannelTypeXunfei {
+		c.Set("base_url", channel.GetBaseURL())
+		switch channel.Type {
+		case common.ChannelTypeAzure:
 			c.Set("api_version", channel.Other)
+		case common.ChannelTypeXunfei:
+			c.Set("api_version", channel.Other)
+		case common.ChannelTypeGemini:
+			c.Set("api_version", channel.Other)
+		case common.ChannelTypeAIProxyLibrary:
+			c.Set("library_id", channel.Other)
+		case common.ChannelTypeAli:
+			c.Set("plugin", channel.Other)
 		}
 		c.Next()
 	}

middleware/logger.go

@@ -0,0 +1,25 @@
+package middleware
+
+import (
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"one-api/common"
+)
+
+func SetUpLogger(server *gin.Engine) {
+	server.Use(gin.LoggerWithFormatter(func(param gin.LogFormatterParams) string {
+		var requestID string
+		if param.Keys != nil {
+			requestID = param.Keys[common.RequestIdKey].(string)
+		}
+		return fmt.Sprintf("[GIN] %s | %s | %3d | %13v | %15s | %7s %s\n",
+			param.TimeStamp.Format("2006/01/02 - 15:04:05"),
+			requestID,
+			param.StatusCode,
+			param.Latency,
+			param.ClientIP,
+			param.Method,
+			param.Path,
+		)
+	}))
+}

middleware/recover.go

@@ -0,0 +1,26 @@
+package middleware
+
+import (
+	"fmt"
+	"github.com/gin-gonic/gin"
+	"net/http"
+	"one-api/common"
+)
+
+func RelayPanicRecover() gin.HandlerFunc {
+	return func(c *gin.Context) {
+		defer func() {
+			if err := recover(); err != nil {
+				common.SysError(fmt.Sprintf("panic detected: %v", err))
+				c.JSON(http.StatusInternalServerError, gin.H{
+					"error": gin.H{
+						"message": fmt.Sprintf("Panic detected, error: %v. Please submit a issue here: https://github.com/songquanpeng/one-api", err),
+						"type":    "one_api_panic",
+					},
+				})
+				c.Abort()
+			}
+		}()
+		c.Next()
+	}
+}

middleware/request-id.go

@@ -0,0 +1,18 @@
+package middleware
+
+import (
+	"context"
+	"github.com/gin-gonic/gin"
+	"one-api/common"
+)
+
+func RequestId() func(c *gin.Context) {
+	return func(c *gin.Context) {
+		id := common.GetTimeString() + common.GetRandomString(8)
+		c.Set(common.RequestIdKey, id)
+		ctx := context.WithValue(c.Request.Context(), common.RequestIdKey, id)
+		c.Request = c.Request.WithContext(ctx)
+		c.Header(common.RequestIdKey, id)
+		c.Next()
+	}
+}

middleware/utils.go

@@ -0,0 +1,17 @@
+package middleware
+
+import (
+	"github.com/gin-gonic/gin"
+	"one-api/common"
+)
+
+func abortWithMessage(c *gin.Context, statusCode int, message string) {
+	c.JSON(statusCode, gin.H{
+		"error": gin.H{
+			"message": common.MessageWithRequestId(message, c.GetString(common.RequestIdKey)),
+			"type":    "one_api_error",
+		},
+	})
+	c.Abort()
+	common.LogError(c.Request.Context(), message)
+}

model/ability.go

@@ -10,15 +10,25 @@ type Ability struct {
 	Model     string `json:"model" gorm:"primaryKey;autoIncrement:false"`
 	ChannelId int    `json:"channel_id" gorm:"primaryKey;autoIncrement:false;index"`
 	Enabled   bool   `json:"enabled"`
+	Priority  *int64 `json:"priority" gorm:"bigint;default:0;index"`
 }
 
 func GetRandomSatisfiedChannel(group string, model string) (*Channel, error) {
 	ability := Ability{}
+	groupCol := "`group`"
+	trueVal := "1"
+	if common.UsingPostgreSQL {
+		groupCol = `"group"`
+		trueVal = "true"
+	}
+
 	var err error = nil
-	if common.UsingSQLite {
-		err = DB.Where("`group` = ? and model = ? and enabled = 1", group, model).Order("RANDOM()").Limit(1).First(&ability).Error
+	maxPrioritySubQuery := DB.Model(&Ability{}).Select("MAX(priority)").Where(groupCol+" = ? and model = ? and enabled = "+trueVal, group, model)
+	channelQuery := DB.Where(groupCol+" = ? and model = ? and enabled = "+trueVal+" and priority = (?)", group, model, maxPrioritySubQuery)
+	if common.UsingSQLite || common.UsingPostgreSQL {
+		err = channelQuery.Order("RANDOM()").First(&ability).Error
 	} else {
-		err = DB.Where("`group` = ? and model = ? and enabled = 1", group, model).Order("RAND()").Limit(1).First(&ability).Error
+		err = channelQuery.Order("RAND()").First(&ability).Error
 	}
 	if err != nil {
 		return nil, err
@@ -40,6 +50,7 @@ func (channel *Channel) AddAbilities() error {
 				Model:     model,
 				ChannelId: channel.Id,
 				Enabled:   channel.Status == common.ChannelStatusEnabled,
+				Priority:  channel.Priority,
 			}
 			abilities = append(abilities, ability)
 		}

model/cache.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"math/rand"
 	"one-api/common"
+	"sort"
 	"strconv"
 	"strings"
 	"sync"
@@ -20,14 +21,18 @@ var (
 )
 
 func CacheGetTokenByKey(key string) (*Token, error) {
+	keyCol := "`key`"
+	if common.UsingPostgreSQL {
+		keyCol = `"key"`
+	}
 	var token Token
 	if !common.RedisEnabled {
-		err := DB.Where("`key` = ?", key).First(&token).Error
+		err := DB.Where(keyCol+" = ?", key).First(&token).Error
 		return &token, err
 	}
 	tokenObjectString, err := common.RedisGet(fmt.Sprintf("token:%s", key))
 	if err != nil {
-		err := DB.Where("`key` = ?", key).First(&token).Error
+		err := DB.Where(keyCol+" = ?", key).First(&token).Error
 		if err != nil {
 			return nil, err
 		}
@@ -103,23 +108,28 @@ func CacheDecreaseUserQuota(id int, quota int) error {
 	return err
 }
 
-func CacheIsUserEnabled(userId int) bool {
+func CacheIsUserEnabled(userId int) (bool, error) {
 	if !common.RedisEnabled {
 		return IsUserEnabled(userId)
 	}
 	enabled, err := common.RedisGet(fmt.Sprintf("user_enabled:%d", userId))
-	if err != nil {
-		status := common.UserStatusDisabled
-		if IsUserEnabled(userId) {
-			status = common.UserStatusEnabled
-		}
-		enabled = fmt.Sprintf("%d", status)
-		err = common.RedisSet(fmt.Sprintf("user_enabled:%d", userId), enabled, time.Duration(UserId2StatusCacheSeconds)*time.Second)
-		if err != nil {
-			common.SysError("Redis set user enabled error: " + err.Error())
-		}
+	if err == nil {
+		return enabled == "1", nil
 	}
-	return enabled == "1"
+
+	userEnabled, err := IsUserEnabled(userId)
+	if err != nil {
+		return false, err
+	}
+	enabled = "0"
+	if userEnabled {
+		enabled = "1"
+	}
+	err = common.RedisSet(fmt.Sprintf("user_enabled:%d", userId), enabled, time.Duration(UserId2StatusCacheSeconds)*time.Second)
+	if err != nil {
+		common.SysError("Redis set user enabled error: " + err.Error())
+	}
+	return userEnabled, err
 }
 
 var group2model2channels map[string]map[string][]*Channel
@@ -154,6 +164,17 @@ func InitChannelCache() {
 			}
 		}
 	}
+
+	// sort by priority
+	for group, model2channels := range newGroup2model2channels {
+		for model, channels := range model2channels {
+			sort.Slice(channels, func(i, j int) bool {
+				return channels[i].GetPriority() > channels[j].GetPriority()
+			})
+			newGroup2model2channels[group][model] = channels
+		}
+	}
+
 	channelSyncLock.Lock()
 	group2model2channels = newGroup2model2channels
 	channelSyncLock.Unlock()
@@ -169,7 +190,7 @@ func SyncChannelCache(frequency int) {
 }
 
 func CacheGetRandomSatisfiedChannel(group string, model string) (*Channel, error) {
-	if !common.RedisEnabled {
+	if !common.MemoryCacheEnabled {
 		return GetRandomSatisfiedChannel(group, model)
 	}
 	channelSyncLock.RLock()
@@ -178,6 +199,17 @@ func CacheGetRandomSatisfiedChannel(group string, model string) (*Channel, error
 	if len(channels) == 0 {
 		return nil, errors.New("channel not found")
 	}
-	idx := rand.Intn(len(channels))
+	endIdx := len(channels)
+	// choose by priority
+	firstChannel := channels[0]
+	if firstChannel.GetPriority() > 0 {
+		for i := range channels {
+			if channels[i].GetPriority() != firstChannel.GetPriority() {
+				endIdx = i
+				break
+			}
+		}
+	}
+	idx := rand.Intn(endIdx)
 	return channels[idx], nil
 }

model/channel.go

@@ -11,18 +11,19 @@ type Channel struct {
 	Key                string  `json:"key" gorm:"not null;index"`
 	Status             int     `json:"status" gorm:"default:1"`
 	Name               string  `json:"name" gorm:"index"`
-	Weight             int     `json:"weight"`
+	Weight             *uint   `json:"weight" gorm:"default:0"`
 	CreatedTime        int64   `json:"created_time" gorm:"bigint"`
 	TestTime           int64   `json:"test_time" gorm:"bigint"`
 	ResponseTime       int     `json:"response_time"` // in milliseconds
-	BaseURL            string  `json:"base_url" gorm:"column:base_url"`
+	BaseURL            *string `json:"base_url" gorm:"column:base_url;default:''"`
 	Other              string  `json:"other"`
 	Balance            float64 `json:"balance"` // in USD
 	BalanceUpdatedTime int64   `json:"balance_updated_time" gorm:"bigint"`
 	Models             string  `json:"models"`
 	Group              string  `json:"group" gorm:"type:varchar(32);default:'default'"`
 	UsedQuota          int64   `json:"used_quota" gorm:"bigint;default:0"`
-	ModelMapping       string  `json:"model_mapping" gorm:"type:varchar(1024);default:''"`
+	ModelMapping       *string `json:"model_mapping" gorm:"type:varchar(1024);default:''"`
+	Priority           *int64  `json:"priority" gorm:"bigint;default:0"`
 }
 
 func GetAllChannels(startIdx int, num int, selectAll bool) ([]*Channel, error) {
@@ -37,7 +38,11 @@ func GetAllChannels(startIdx int, num int, selectAll bool) ([]*Channel, error) {
 }
 
 func SearchChannels(keyword string) (channels []*Channel, err error) {
-	err = DB.Omit("key").Where("id = ? or name LIKE ? or `key` = ?", keyword, keyword+"%", keyword).Find(&channels).Error
+	keyCol := "`key`"
+	if common.UsingPostgreSQL {
+		keyCol = `"key"`
+	}
+	err = DB.Omit("key").Where("id = ? or name LIKE ? or "+keyCol+" = ?", common.String2Int(keyword), keyword+"%", keyword).Find(&channels).Error
 	return channels, err
 }
 
@@ -52,17 +57,6 @@ func GetChannelById(id int, selectAll bool) (*Channel, error) {
 	return &channel, err
 }
 
-func GetRandomChannel() (*Channel, error) {
-	channel := Channel{}
-	var err error = nil
-	if common.UsingSQLite {
-		err = DB.Where("status = ? and `group` = ?", common.ChannelStatusEnabled, "default").Order("RANDOM()").Limit(1).First(&channel).Error
-	} else {
-		err = DB.Where("status = ? and `group` = ?", common.ChannelStatusEnabled, "default").Order("RAND()").Limit(1).First(&channel).Error
-	}
-	return &channel, err
-}
-
 func BatchInsertChannels(channels []Channel) error {
 	var err error
 	err = DB.Create(&channels).Error
@@ -78,6 +72,27 @@ func BatchInsertChannels(channels []Channel) error {
 	return nil
 }
 
+func (channel *Channel) GetPriority() int64 {
+	if channel.Priority == nil {
+		return 0
+	}
+	return *channel.Priority
+}
+
+func (channel *Channel) GetBaseURL() string {
+	if channel.BaseURL == nil {
+		return ""
+	}
+	return *channel.BaseURL
+}
+
+func (channel *Channel) GetModelMapping() string {
+	if channel.ModelMapping == nil {
+		return ""
+	}
+	return *channel.ModelMapping
+}
+
 func (channel *Channel) Insert() error {
 	var err error
 	err = DB.Create(channel).Error
@@ -141,8 +156,26 @@ func UpdateChannelStatusById(id int, status int) {
 }
 
 func UpdateChannelUsedQuota(id int, quota int) {
+	if common.BatchUpdateEnabled {
+		addNewRecord(BatchUpdateTypeChannelUsedQuota, id, quota)
+		return
+	}
+	updateChannelUsedQuota(id, quota)
+}
+
+func updateChannelUsedQuota(id int, quota int) {
 	err := DB.Model(&Channel{}).Where("id = ?", id).Update("used_quota", gorm.Expr("used_quota + ?", quota)).Error
 	if err != nil {
 		common.SysError("failed to update channel used quota: " + err.Error())
 	}
 }
+
+func DeleteChannelByStatus(status int64) (int64, error) {
+	result := DB.Where("status = ?", status).Delete(&Channel{})
+	return result.RowsAffected, result.Error
+}
+
+func DeleteDisabledChannel() (int64, error) {
+	result := DB.Where("status = ? or status = ?", common.ChannelStatusAutoDisabled, common.ChannelStatusManuallyDisabled).Delete(&Channel{})
+	return result.RowsAffected, result.Error
+}

model/log.go

@@ -1,22 +1,25 @@
 package model
 
 import (
+	"context"
+	"fmt"
 	"gorm.io/gorm"
 	"one-api/common"
 )
 
 type Log struct {
-	Id               int    `json:"id"`
-	UserId           int    `json:"user_id"`
-	CreatedAt        int64  `json:"created_at" gorm:"bigint;index"`
-	Type             int    `json:"type" gorm:"index"`
+	Id               int    `json:"id;index:idx_created_at_id,priority:1"`
+	UserId           int    `json:"user_id" gorm:"index"`
+	CreatedAt        int64  `json:"created_at" gorm:"bigint;index:idx_created_at_id,priority:2;index:idx_created_at_type"`
+	Type             int    `json:"type" gorm:"index:idx_created_at_type"`
 	Content          string `json:"content"`
-	Username         string `json:"username" gorm:"index;default:''"`
+	Username         string `json:"username" gorm:"index:index_username_model_name,priority:2;default:''"`
 	TokenName        string `json:"token_name" gorm:"index;default:''"`
-	ModelName        string `json:"model_name" gorm:"index;default:''"`
+	ModelName        string `json:"model_name" gorm:"index;index:index_username_model_name,priority:1;default:''"`
 	Quota            int    `json:"quota" gorm:"default:0"`
 	PromptTokens     int    `json:"prompt_tokens" gorm:"default:0"`
 	CompletionTokens int    `json:"completion_tokens" gorm:"default:0"`
+	ChannelId        int    `json:"channel" gorm:"index"`
 }
 
 const (
@@ -44,7 +47,8 @@ func RecordLog(userId int, logType int, content string) {
 	}
 }
 
-func RecordConsumeLog(userId int, promptTokens int, completionTokens int, modelName string, tokenName string, quota int, content string) {
+func RecordConsumeLog(ctx context.Context, userId int, channelId int, promptTokens int, completionTokens int, modelName string, tokenName string, quota int, content string) {
+	common.LogInfo(ctx, fmt.Sprintf("record consume log: userId=%d, channelId=%d, promptTokens=%d, completionTokens=%d, modelName=%s, tokenName=%s, quota=%d, content=%s", userId, channelId, promptTokens, completionTokens, modelName, tokenName, quota, content))
 	if !common.LogConsumeEnabled {
 		return
 	}
@@ -59,14 +63,15 @@ func RecordConsumeLog(userId int, promptTokens int, completionTokens int, modelN
 		TokenName:        tokenName,
 		ModelName:        modelName,
 		Quota:            quota,
+		ChannelId:        channelId,
 	}
 	err := DB.Create(log).Error
 	if err != nil {
-		common.SysError("failed to record log: " + err.Error())
+		common.LogError(ctx, "failed to record log: "+err.Error())
 	}
 }
 
-func GetAllLogs(logType int, startTimestamp int64, endTimestamp int64, modelName string, username string, tokenName string, startIdx int, num int) (logs []*Log, err error) {
+func GetAllLogs(logType int, startTimestamp int64, endTimestamp int64, modelName string, username string, tokenName string, startIdx int, num int, channel int) (logs []*Log, err error) {
 	var tx *gorm.DB
 	if logType == LogTypeUnknown {
 		tx = DB
@@ -88,6 +93,9 @@ func GetAllLogs(logType int, startTimestamp int64, endTimestamp int64, modelName
 	if endTimestamp != 0 {
 		tx = tx.Where("created_at <= ?", endTimestamp)
 	}
+	if channel != 0 {
+		tx = tx.Where("channel_id = ?", channel)
+	}
 	err = tx.Order("id desc").Limit(num).Offset(startIdx).Find(&logs).Error
 	return logs, err
 }
@@ -125,8 +133,8 @@ func SearchUserLogs(userId int, keyword string) (logs []*Log, err error) {
 	return logs, err
 }
 
-func SumUsedQuota(logType int, startTimestamp int64, endTimestamp int64, modelName string, username string, tokenName string) (quota int) {
-	tx := DB.Table("logs").Select("sum(quota)")
+func SumUsedQuota(logType int, startTimestamp int64, endTimestamp int64, modelName string, username string, tokenName string, channel int) (quota int) {
+	tx := DB.Table("logs").Select("ifnull(sum(quota),0)")
 	if username != "" {
 		tx = tx.Where("username = ?", username)
 	}
@@ -142,12 +150,15 @@ func SumUsedQuota(logType int, startTimestamp int64, endTimestamp int64, modelNa
 	if modelName != "" {
 		tx = tx.Where("model_name = ?", modelName)
 	}
+	if channel != 0 {
+		tx = tx.Where("channel_id = ?", channel)
+	}
 	tx.Where("type = ?", LogTypeConsume).Scan(&quota)
 	return quota
 }
 
 func SumUsedToken(logType int, startTimestamp int64, endTimestamp int64, modelName string, username string, tokenName string) (token int) {
-	tx := DB.Table("logs").Select("sum(prompt_tokens) + sum(completion_tokens)")
+	tx := DB.Table("logs").Select("ifnull(sum(prompt_tokens),0) + ifnull(sum(completion_tokens),0)")
 	if username != "" {
 		tx = tx.Where("username = ?", username)
 	}
@@ -166,3 +177,8 @@ func SumUsedToken(logType int, startTimestamp int64, endTimestamp int64, modelNa
 	tx.Where("type = ?", LogTypeConsume).Scan(&token)
 	return token
 }
+
+func DeleteOldLog(targetTimestamp int64) (int64, error) {
+	result := DB.Where("created_at < ?", targetTimestamp).Delete(&Log{})
+	return result.RowsAffected, result.Error
+}

model/main.go

@@ -42,6 +42,7 @@ func chooseDB() (*gorm.DB, error) {
 		if strings.HasPrefix(dsn, "postgres://") {
 			// Use PostgreSQL
 			common.SysLog("using PostgreSQL as database")
+			common.UsingPostgreSQL = true
 			return gorm.Open(postgres.New(postgres.Config{
 				DSN:                  dsn,
 				PreferSimpleProtocol: true, // disables implicit prepared statement usage
@@ -81,6 +82,7 @@ func InitDB() (err error) {
 		if !common.IsMasterNode {
 			return nil
 		}
+		common.SysLog("database migration started")
 		err = db.AutoMigrate(&Channel{})
 		if err != nil {
 			return err

model/option.go

@@ -34,6 +34,7 @@ func InitOptionMap() {
 	common.OptionMap["TurnstileCheckEnabled"] = strconv.FormatBool(common.TurnstileCheckEnabled)
 	common.OptionMap["RegisterEnabled"] = strconv.FormatBool(common.RegisterEnabled)
 	common.OptionMap["AutomaticDisableChannelEnabled"] = strconv.FormatBool(common.AutomaticDisableChannelEnabled)
+	common.OptionMap["AutomaticEnableChannelEnabled"] = strconv.FormatBool(common.AutomaticEnableChannelEnabled)
 	common.OptionMap["ApproximateTokenEnabled"] = strconv.FormatBool(common.ApproximateTokenEnabled)
 	common.OptionMap["LogConsumeEnabled"] = strconv.FormatBool(common.LogConsumeEnabled)
 	common.OptionMap["DisplayInCurrencyEnabled"] = strconv.FormatBool(common.DisplayInCurrencyEnabled)
@@ -147,6 +148,8 @@ func updateOptionMap(key string, value string) (err error) {
 			common.EmailDomainRestrictionEnabled = boolValue
 		case "AutomaticDisableChannelEnabled":
 			common.AutomaticDisableChannelEnabled = boolValue
+		case "AutomaticEnableChannelEnabled":
+			common.AutomaticEnableChannelEnabled = boolValue
 		case "ApproximateTokenEnabled":
 			common.ApproximateTokenEnabled = boolValue
 		case "LogConsumeEnabled":

model/redemption.go

@@ -50,8 +50,13 @@ func Redeem(key string, userId int) (quota int, err error) {
 	}
 	redemption := &Redemption{}
 
+	keyCol := "`key`"
+	if common.UsingPostgreSQL {
+		keyCol = `"key"`
+	}
+
 	err = DB.Transaction(func(tx *gorm.DB) error {
-		err := tx.Set("gorm:query_option", "FOR UPDATE").Where("`key` = ?", key).First(redemption).Error
+		err := tx.Set("gorm:query_option", "FOR UPDATE").Where(keyCol+" = ?", key).First(redemption).Error
 		if err != nil {
 			return errors.New("无效的兑换码")
 		}

model/token.go

@@ -39,32 +39,35 @@ func ValidateUserToken(key string) (token *Token, err error) {
 	}
 	token, err = CacheGetTokenByKey(key)
 	if err == nil {
+		if token.Status == common.TokenStatusExhausted {
+			return nil, errors.New("该令牌额度已用尽")
+		} else if token.Status == common.TokenStatusExpired {
+			return nil, errors.New("该令牌已过期")
+		}
 		if token.Status != common.TokenStatusEnabled {
 			return nil, errors.New("该令牌状态不可用")
 		}
 		if token.ExpiredTime != -1 && token.ExpiredTime < common.GetTimestamp() {
-			token.Status = common.TokenStatusExpired
-			err := token.SelectUpdate()
-			if err != nil {
-				common.SysError("failed to update token status" + err.Error())
+			if !common.RedisEnabled {
+				token.Status = common.TokenStatusExpired
+				err := token.SelectUpdate()
+				if err != nil {
+					common.SysError("failed to update token status" + err.Error())
+				}
 			}
 			return nil, errors.New("该令牌已过期")
 		}
 		if !token.UnlimitedQuota && token.RemainQuota <= 0 {
-			token.Status = common.TokenStatusExhausted
-			err := token.SelectUpdate()
-			if err != nil {
-				common.SysError("failed to update token status" + err.Error())
+			if !common.RedisEnabled {
+				// in this case, we can make sure the token is exhausted
+				token.Status = common.TokenStatusExhausted
+				err := token.SelectUpdate()
+				if err != nil {
+					common.SysError("failed to update token status" + err.Error())
+				}
 			}
 			return nil, errors.New("该令牌额度已用尽")
 		}
-		go func() {
-			token.AccessedTime = common.GetTimestamp()
-			err := token.SelectUpdate()
-			if err != nil {
-				common.SysError("failed to update token" + err.Error())
-			}
-		}()
 		return token, nil
 	}
 	return nil, errors.New("无效的令牌")
@@ -131,10 +134,19 @@ func IncreaseTokenQuota(id int, quota int) (err error) {
 	if quota < 0 {
 		return errors.New("quota 不能为负数！")
 	}
+	if common.BatchUpdateEnabled {
+		addNewRecord(BatchUpdateTypeTokenQuota, id, quota)
+		return nil
+	}
+	return increaseTokenQuota(id, quota)
+}
+
+func increaseTokenQuota(id int, quota int) (err error) {
 	err = DB.Model(&Token{}).Where("id = ?", id).Updates(
 		map[string]interface{}{
-			"remain_quota": gorm.Expr("remain_quota + ?", quota),
-			"used_quota":   gorm.Expr("used_quota - ?", quota),
+			"remain_quota":  gorm.Expr("remain_quota + ?", quota),
+			"used_quota":    gorm.Expr("used_quota - ?", quota),
+			"accessed_time": common.GetTimestamp(),
 		},
 	).Error
 	return err
@@ -144,10 +156,19 @@ func DecreaseTokenQuota(id int, quota int) (err error) {
 	if quota < 0 {
 		return errors.New("quota 不能为负数！")
 	}
+	if common.BatchUpdateEnabled {
+		addNewRecord(BatchUpdateTypeTokenQuota, id, -quota)
+		return nil
+	}
+	return decreaseTokenQuota(id, quota)
+}
+
+func decreaseTokenQuota(id int, quota int) (err error) {
 	err = DB.Model(&Token{}).Where("id = ?", id).Updates(
 		map[string]interface{}{
-			"remain_quota": gorm.Expr("remain_quota - ?", quota),
-			"used_quota":   gorm.Expr("used_quota + ?", quota),
+			"remain_quota":  gorm.Expr("remain_quota - ?", quota),
+			"used_quota":    gorm.Expr("used_quota + ?", quota),
+			"accessed_time": common.GetTimestamp(),
 		},
 	).Error
 	return err

model/user.go

@@ -226,17 +226,16 @@ func IsAdmin(userId int) bool {
 	return user.Role >= common.RoleAdminUser
 }
 
-func IsUserEnabled(userId int) bool {
+func IsUserEnabled(userId int) (bool, error) {
 	if userId == 0 {
-		return false
+		return false, errors.New("user id is empty")
 	}
 	var user User
 	err := DB.Where("id = ?", userId).Select("status").Find(&user).Error
 	if err != nil {
-		common.SysError("no such user " + err.Error())
-		return false
+		return false, err
 	}
-	return user.Status == common.UserStatusEnabled
+	return user.Status == common.UserStatusEnabled, nil
 }
 
 func ValidateAccessToken(token string) (user *User) {
@@ -267,7 +266,12 @@ func GetUserEmail(id int) (email string, err error) {
 }
 
 func GetUserGroup(id int) (group string, err error) {
-	err = DB.Model(&User{}).Where("id = ?", id).Select("`group`").Find(&group).Error
+	groupCol := "`group`"
+	if common.UsingPostgreSQL {
+		groupCol = `"group"`
+	}
+
+	err = DB.Model(&User{}).Where("id = ?", id).Select(groupCol).Find(&group).Error
 	return group, err
 }
 
@@ -275,6 +279,14 @@ func IncreaseUserQuota(id int, quota int) (err error) {
 	if quota < 0 {
 		return errors.New("quota 不能为负数！")
 	}
+	if common.BatchUpdateEnabled {
+		addNewRecord(BatchUpdateTypeUserQuota, id, quota)
+		return nil
+	}
+	return increaseUserQuota(id, quota)
+}
+
+func increaseUserQuota(id int, quota int) (err error) {
 	err = DB.Model(&User{}).Where("id = ?", id).Update("quota", gorm.Expr("quota + ?", quota)).Error
 	return err
 }
@@ -283,6 +295,14 @@ func DecreaseUserQuota(id int, quota int) (err error) {
 	if quota < 0 {
 		return errors.New("quota 不能为负数！")
 	}
+	if common.BatchUpdateEnabled {
+		addNewRecord(BatchUpdateTypeUserQuota, id, -quota)
+		return nil
+	}
+	return decreaseUserQuota(id, quota)
+}
+
+func decreaseUserQuota(id int, quota int) (err error) {
 	err = DB.Model(&User{}).Where("id = ?", id).Update("quota", gorm.Expr("quota - ?", quota)).Error
 	return err
 }
@@ -293,10 +313,19 @@ func GetRootUserEmail() (email string) {
 }
 
 func UpdateUserUsedQuotaAndRequestCount(id int, quota int) {
+	if common.BatchUpdateEnabled {
+		addNewRecord(BatchUpdateTypeUsedQuota, id, quota)
+		addNewRecord(BatchUpdateTypeRequestCount, id, 1)
+		return
+	}
+	updateUserUsedQuotaAndRequestCount(id, quota, 1)
+}
+
+func updateUserUsedQuotaAndRequestCount(id int, quota int, count int) {
 	err := DB.Model(&User{}).Where("id = ?", id).Updates(
 		map[string]interface{}{
 			"used_quota":    gorm.Expr("used_quota + ?", quota),
-			"request_count": gorm.Expr("request_count + ?", 1),
+			"request_count": gorm.Expr("request_count + ?", count),
 		},
 	).Error
 	if err != nil {
@@ -304,6 +333,24 @@ func UpdateUserUsedQuotaAndRequestCount(id int, quota int) {
 	}
 }
 
+func updateUserUsedQuota(id int, quota int) {
+	err := DB.Model(&User{}).Where("id = ?", id).Updates(
+		map[string]interface{}{
+			"used_quota": gorm.Expr("used_quota + ?", quota),
+		},
+	).Error
+	if err != nil {
+		common.SysError("failed to update user used quota: " + err.Error())
+	}
+}
+
+func updateUserRequestCount(id int, count int) {
+	err := DB.Model(&User{}).Where("id = ?", id).Update("request_count", gorm.Expr("request_count + ?", count)).Error
+	if err != nil {
+		common.SysError("failed to update user request count: " + err.Error())
+	}
+}
+
 func GetUsernameById(id int) (username string) {
 	DB.Model(&User{}).Where("id = ?", id).Select("username").Find(&username)
 	return username

model/utils.go

@@ -0,0 +1,77 @@
+package model
+
+import (
+	"one-api/common"
+	"sync"
+	"time"
+)
+
+const (
+	BatchUpdateTypeUserQuota = iota
+	BatchUpdateTypeTokenQuota
+	BatchUpdateTypeUsedQuota
+	BatchUpdateTypeChannelUsedQuota
+	BatchUpdateTypeRequestCount
+	BatchUpdateTypeCount // if you add a new type, you need to add a new map and a new lock
+)
+
+var batchUpdateStores []map[int]int
+var batchUpdateLocks []sync.Mutex
+
+func init() {
+	for i := 0; i < BatchUpdateTypeCount; i++ {
+		batchUpdateStores = append(batchUpdateStores, make(map[int]int))
+		batchUpdateLocks = append(batchUpdateLocks, sync.Mutex{})
+	}
+}
+
+func InitBatchUpdater() {
+	go func() {
+		for {
+			time.Sleep(time.Duration(common.BatchUpdateInterval) * time.Second)
+			batchUpdate()
+		}
+	}()
+}
+
+func addNewRecord(type_ int, id int, value int) {
+	batchUpdateLocks[type_].Lock()
+	defer batchUpdateLocks[type_].Unlock()
+	if _, ok := batchUpdateStores[type_][id]; !ok {
+		batchUpdateStores[type_][id] = value
+	} else {
+		batchUpdateStores[type_][id] += value
+	}
+}
+
+func batchUpdate() {
+	common.SysLog("batch update started")
+	for i := 0; i < BatchUpdateTypeCount; i++ {
+		batchUpdateLocks[i].Lock()
+		store := batchUpdateStores[i]
+		batchUpdateStores[i] = make(map[int]int)
+		batchUpdateLocks[i].Unlock()
+		// TODO: maybe we can combine updates with same key?
+		for key, value := range store {
+			switch i {
+			case BatchUpdateTypeUserQuota:
+				err := increaseUserQuota(key, value)
+				if err != nil {
+					common.SysError("failed to batch update user quota: " + err.Error())
+				}
+			case BatchUpdateTypeTokenQuota:
+				err := increaseTokenQuota(key, value)
+				if err != nil {
+					common.SysError("failed to batch update token quota: " + err.Error())
+				}
+			case BatchUpdateTypeUsedQuota:
+				updateUserUsedQuota(key, value)
+			case BatchUpdateTypeRequestCount:
+				updateUserRequestCount(key, value)
+			case BatchUpdateTypeChannelUsedQuota:
+				updateChannelUsedQuota(key, value)
+			}
+		}
+	}
+	common.SysLog("batch update finished")
+}

pull_request_template.md

@@ -0,0 +1,3 @@
+close #issue_number
+
+我已确认该 PR 已自测通过，相关截图如下：

router/api-router.go

@@ -21,6 +21,7 @@ func SetApiRouter(router *gin.Engine) {
 		apiRouter.GET("/reset_password", middleware.CriticalRateLimit(), middleware.TurnstileCheck(), controller.SendPasswordResetEmail)
 		apiRouter.POST("/user/reset", middleware.CriticalRateLimit(), controller.ResetPassword)
 		apiRouter.GET("/oauth/github", middleware.CriticalRateLimit(), controller.GitHubOAuth)
+		apiRouter.GET("/oauth/state", middleware.CriticalRateLimit(), controller.GenerateOAuthCode)
 		apiRouter.GET("/oauth/wechat", middleware.CriticalRateLimit(), controller.WeChatAuth)
 		apiRouter.GET("/oauth/wechat/bind", middleware.CriticalRateLimit(), middleware.UserAuth(), controller.WeChatBind)
 		apiRouter.GET("/oauth/email/bind", middleware.CriticalRateLimit(), middleware.UserAuth(), controller.EmailBind)
@@ -73,6 +74,7 @@ func SetApiRouter(router *gin.Engine) {
 			channelRoute.GET("/update_balance/:id", controller.UpdateChannelBalance)
 			channelRoute.POST("/", controller.AddChannel)
 			channelRoute.PUT("/", controller.UpdateChannel)
+			channelRoute.DELETE("/disabled", controller.DeleteDisabledChannel)
 			channelRoute.DELETE("/:id", controller.DeleteChannel)
 		}
 		tokenRoute := apiRouter.Group("/token")
@@ -97,6 +99,7 @@ func SetApiRouter(router *gin.Engine) {
 		}
 		logRoute := apiRouter.Group("/log")
 		logRoute.GET("/", middleware.AdminAuth(), controller.GetAllLogs)
+		logRoute.DELETE("/", middleware.AdminAuth(), controller.DeleteHistoryLogs)
 		logRoute.GET("/stat", middleware.AdminAuth(), controller.GetLogsStat)
 		logRoute.GET("/self/stat", middleware.UserAuth(), controller.GetLogsSelfStat)
 		logRoute.GET("/search", middleware.AdminAuth(), controller.SearchAllLogs)

router/relay-router.go

@@ -8,6 +8,7 @@ import (
 )
 
 func SetRelayRouter(router *gin.Engine) {
+	router.Use(middleware.CORS())
 	// https://platform.openai.com/docs/api-reference/introduction
 	modelsRouter := router.Group("/v1/models")
 	modelsRouter.Use(middleware.TokenAuth())
@@ -16,7 +17,7 @@ func SetRelayRouter(router *gin.Engine) {
 		modelsRouter.GET("/:model", controller.RetrieveModel)
 	}
 	relayV1Router := router.Group("/v1")
-	relayV1Router.Use(middleware.TokenAuth(), middleware.Distribute())
+	relayV1Router.Use(middleware.RelayPanicRecover(), middleware.TokenAuth(), middleware.Distribute())
 	{
 		relayV1Router.POST("/completions", controller.Relay)
 		relayV1Router.POST("/chat/completions", controller.Relay)
@@ -26,19 +27,46 @@ func SetRelayRouter(router *gin.Engine) {
 		relayV1Router.POST("/images/variations", controller.RelayNotImplemented)
 		relayV1Router.POST("/embeddings", controller.Relay)
 		relayV1Router.POST("/engines/:model/embeddings", controller.Relay)
-		relayV1Router.POST("/audio/transcriptions", controller.RelayNotImplemented)
-		relayV1Router.POST("/audio/translations", controller.RelayNotImplemented)
+		relayV1Router.POST("/audio/transcriptions", controller.Relay)
+		relayV1Router.POST("/audio/translations", controller.Relay)
+		relayV1Router.POST("/audio/speech", controller.Relay)
 		relayV1Router.GET("/files", controller.RelayNotImplemented)
 		relayV1Router.POST("/files", controller.RelayNotImplemented)
 		relayV1Router.DELETE("/files/:id", controller.RelayNotImplemented)
 		relayV1Router.GET("/files/:id", controller.RelayNotImplemented)
 		relayV1Router.GET("/files/:id/content", controller.RelayNotImplemented)
-		relayV1Router.POST("/fine-tunes", controller.RelayNotImplemented)
-		relayV1Router.GET("/fine-tunes", controller.RelayNotImplemented)
-		relayV1Router.GET("/fine-tunes/:id", controller.RelayNotImplemented)
-		relayV1Router.POST("/fine-tunes/:id/cancel", controller.RelayNotImplemented)
-		relayV1Router.GET("/fine-tunes/:id/events", controller.RelayNotImplemented)
+		relayV1Router.POST("/fine_tuning/jobs", controller.RelayNotImplemented)
+		relayV1Router.GET("/fine_tuning/jobs", controller.RelayNotImplemented)
+		relayV1Router.GET("/fine_tuning/jobs/:id", controller.RelayNotImplemented)
+		relayV1Router.POST("/fine_tuning/jobs/:id/cancel", controller.RelayNotImplemented)
+		relayV1Router.GET("/fine_tuning/jobs/:id/events", controller.RelayNotImplemented)
 		relayV1Router.DELETE("/models/:model", controller.RelayNotImplemented)
 		relayV1Router.POST("/moderations", controller.Relay)
+		relayV1Router.POST("/assistants", controller.RelayNotImplemented)
+		relayV1Router.GET("/assistants/:id", controller.RelayNotImplemented)
+		relayV1Router.POST("/assistants/:id", controller.RelayNotImplemented)
+		relayV1Router.DELETE("/assistants/:id", controller.RelayNotImplemented)
+		relayV1Router.GET("/assistants", controller.RelayNotImplemented)
+		relayV1Router.POST("/assistants/:id/files", controller.RelayNotImplemented)
+		relayV1Router.GET("/assistants/:id/files/:fileId", controller.RelayNotImplemented)
+		relayV1Router.DELETE("/assistants/:id/files/:fileId", controller.RelayNotImplemented)
+		relayV1Router.GET("/assistants/:id/files", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads/:id", controller.RelayNotImplemented)
+		relayV1Router.DELETE("/threads/:id", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads/:id/messages", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id/messages/:messageId", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads/:id/messages/:messageId", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id/messages/:messageId/files/:filesId", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id/messages/:messageId/files", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads/:id/runs", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id/runs/:runsId", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads/:id/runs/:runsId", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id/runs", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads/:id/runs/:runsId/submit_tool_outputs", controller.RelayNotImplemented)
+		relayV1Router.POST("/threads/:id/runs/:runsId/cancel", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id/runs/:runsId/steps/:stepId", controller.RelayNotImplemented)
+		relayV1Router.GET("/threads/:id/runs/:runsId/steps", controller.RelayNotImplemented)
 	}
 }

web/src/App.js

@@ -283,7 +283,9 @@ function App() {
           </Suspense>
         }
       />
-      <Route path='*' element={NotFound} />
+      <Route path='*' element={
+          <NotFound />
+      } />
     </Routes>
   );
 }

web/src/components/ChannelsTable.js

@@ -1,7 +1,7 @@
 import React, { useEffect, useState } from 'react';
-import { Button, Form, Label, Pagination, Popup, Table } from 'semantic-ui-react';
+import { Button, Form, Input, Label, Message, Pagination, Popup, Table } from 'semantic-ui-react';
 import { Link } from 'react-router-dom';
-import { API, showError, showInfo, showSuccess, timestamp2string } from '../helpers';
+import { API, setPromptShown, shouldShowPrompt, showError, showInfo, showSuccess, timestamp2string } from '../helpers';
 
 import { CHANNEL_OPTIONS, ITEMS_PER_PAGE } from '../constants';
 import { renderGroup, renderNumber } from '../helpers/render';
@@ -24,7 +24,7 @@ function renderType(type) {
     }
     type2label[0] = { value: 0, text: '未知类型', color: 'grey' };
   }
-  return <Label basic color={type2label[type].color}>{type2label[type].text}</Label>;
+  return <Label basic color={type2label[type]?.color}>{type2label[type]?.text}</Label>;
 }
 
 function renderBalance(type, balance) {
@@ -55,6 +55,7 @@ const ChannelsTable = () => {
   const [searchKeyword, setSearchKeyword] = useState('');
   const [searching, setSearching] = useState(false);
   const [updatingBalance, setUpdatingBalance] = useState(false);
+  const [showPrompt, setShowPrompt] = useState(shouldShowPrompt("channel-test"));
 
   const loadChannels = async (startIdx) => {
     const res = await API.get(`/api/channel/?p=${startIdx}`);
@@ -96,7 +97,7 @@ const ChannelsTable = () => {
       });
   }, []);
 
-  const manageChannel = async (id, action, idx) => {
+  const manageChannel = async (id, action, idx, value) => {
     let data = { id };
     let res;
     switch (action) {
@@ -111,6 +112,23 @@ const ChannelsTable = () => {
         data.status = 2;
         res = await API.put('/api/channel/', data);
         break;
+      case 'priority':
+        if (value === '') {
+          return;
+        }
+        data.priority = parseInt(value);
+        res = await API.put('/api/channel/', data);
+        break;
+      case 'weight':
+        if (value === '') {
+          return;
+        }
+        data.weight = parseInt(value);
+        if (data.weight < 0) {
+          data.weight = 0;
+        }
+        res = await API.put('/api/channel/', data);
+        break;
     }
     const { success, message } = res.data;
     if (success) {
@@ -135,9 +153,23 @@ const ChannelsTable = () => {
         return <Label basic color='green'>已启用</Label>;
       case 2:
         return (
-          <Label basic color='red'>
-            已禁用
-          </Label>
+          <Popup
+            trigger={<Label basic color='red'>
+              已禁用
+            </Label>}
+            content='本渠道被手动禁用'
+            basic
+          />
+        );
+      case 3:
+        return (
+          <Popup
+            trigger={<Label basic color='yellow'>
+              已禁用
+            </Label>}
+            content='本渠道被程序自动禁用'
+            basic
+          />
         );
       default:
         return (
@@ -202,7 +234,18 @@ const ChannelsTable = () => {
     const res = await API.get(`/api/channel/test`);
     const { success, message } = res.data;
     if (success) {
-      showInfo('已成功开始测试所有已启用通道，请刷新页面查看结果。');
+      showInfo('已成功开始测试所有通道，请刷新页面查看结果。');
+    } else {
+      showError(message);
+    }
+  };
+
+  const deleteAllDisabledChannels = async () => {
+    const res = await API.delete(`/api/channel/disabled`);
+    const { success, message, data } = res.data;
+    if (success) {
+      showSuccess(`已删除所有禁用渠道，共计 ${data} 个`);
+      await refresh();
     } else {
       showError(message);
     }
@@ -243,17 +286,15 @@ const ChannelsTable = () => {
     if (channels.length === 0) return;
     setLoading(true);
     let sortedChannels = [...channels];
-    if (typeof sortedChannels[0][key] === 'string') {
-      sortedChannels.sort((a, b) => {
+    sortedChannels.sort((a, b) => {
+      if (!isNaN(a[key])) {
+        // If the value is numeric, subtract to sort
+        return a[key] - b[key];
+      } else {
+        // If the value is not numeric, sort as strings
         return ('' + a[key]).localeCompare(b[key]);
-      });
-    } else {
-      sortedChannels.sort((a, b) => {
-        if (a[key] === b[key]) return 0;
-        if (a[key] > b[key]) return -1;
-        if (a[key] < b[key]) return 1;
-      });
-    }
+      }
+    });
     if (sortedChannels[0].id === channels[0].id) {
       sortedChannels.reverse();
     }
@@ -261,6 +302,7 @@ const ChannelsTable = () => {
     setLoading(false);
   };
 
+
   return (
     <>
       <Form onSubmit={searchChannels}>
@@ -274,7 +316,19 @@ const ChannelsTable = () => {
           onChange={handleKeywordChange}
         />
       </Form>
+      {
+        showPrompt && (
+          <Message onDismiss={() => {
+            setShowPrompt(false);
+            setPromptShown("channel-test");
+          }}>
+            当前版本测试是通过按照 OpenAI API 格式使用 gpt-3.5-turbo
+            模型进行非流式请求实现的，因此测试报错并不一定代表通道不可用，该功能后续会修复。
 
+            另外，OpenAI 渠道已经不再支持通过 key 获取余额，因此余额显示为 0。对于支持的渠道类型，请点击余额进行刷新。
+          </Message>
+        )
+      }
       <Table basic compact size='small'>
         <Table.Header>
           <Table.Row>
@@ -334,6 +388,14 @@ const ChannelsTable = () => {
             >
               余额
             </Table.HeaderCell>
+            <Table.HeaderCell
+              style={{ cursor: 'pointer' }}
+              onClick={() => {
+                sortChannel('priority');
+              }}
+            >
+              优先级
+            </Table.HeaderCell>
             <Table.HeaderCell>操作</Table.HeaderCell>
           </Table.Row>
         </Table.Header>
@@ -372,6 +434,22 @@ const ChannelsTable = () => {
                       basic
                     />
                   </Table.Cell>
+                  <Table.Cell>
+                    <Popup
+                      trigger={<Input type='number' defaultValue={channel.priority} onBlur={(event) => {
+                        manageChannel(
+                          channel.id,
+                          'priority',
+                          idx,
+                          event.target.value
+                        );
+                      }}>
+                        <input style={{ maxWidth: '60px' }} />
+                      </Input>}
+                      content='渠道选择优先级，越高越优先'
+                      basic
+                    />
+                  </Table.Cell>
                   <Table.Cell>
                     <div>
                       <Button
@@ -440,7 +518,7 @@ const ChannelsTable = () => {
 
         <Table.Footer>
           <Table.Row>
-            <Table.HeaderCell colSpan='8'>
+            <Table.HeaderCell colSpan='9'>
               <Button size='small' as={Link} to='/channel/add' loading={loading}>
                 添加新的渠道
               </Button>
@@ -449,6 +527,20 @@ const ChannelsTable = () => {
               </Button>
               <Button size='small' onClick={updateAllChannelsBalance}
                       loading={loading || updatingBalance}>更新所有已启用通道余额</Button>
+              <Popup
+                trigger={
+                  <Button size='small' loading={loading}>
+                    删除禁用渠道
+                  </Button>
+                }
+                on='click'
+                flowing
+                hoverable
+              >
+                <Button size='small' loading={loading} negative onClick={deleteAllDisabledChannels}>
+                  确认删除
+                </Button>
+              </Popup>
               <Pagination
                 floated='right'
                 activePage={activePage}

web/src/components/GitHubOAuth.js

@@ -13,8 +13,8 @@ const GitHubOAuth = () => {
 
   let navigate = useNavigate();
 
-  const sendCode = async (code, count) => {
-    const res = await API.get(`/api/oauth/github?code=${code}`);
+  const sendCode = async (code, state, count) => {
+    const res = await API.get(`/api/oauth/github?code=${code}&state=${state}`);
     const { success, message, data } = res.data;
     if (success) {
       if (message === 'bind') {
@@ -36,13 +36,14 @@ const GitHubOAuth = () => {
       count++;
       setPrompt(`出现错误，第 ${count} 次重试中...`);
       await new Promise((resolve) => setTimeout(resolve, count * 2000));
-      await sendCode(code, count);
+      await sendCode(code, state, count);
     }
   };
 
   useEffect(() => {
     let code = searchParams.get('code');
-    sendCode(code, 0).then();
+    let state = searchParams.get('state');
+    sendCode(code, state, 0).then();
   }, []);
 
   return (

web/src/components/LoginForm.js

@@ -2,7 +2,8 @@ import React, { useContext, useEffect, useState } from 'react';
 import { Button, Divider, Form, Grid, Header, Image, Message, Modal, Segment } from 'semantic-ui-react';
 import { Link, useNavigate, useSearchParams } from 'react-router-dom';
 import { UserContext } from '../context/User';
-import { API, getLogo, showError, showSuccess } from '../helpers';
+import { API, getLogo, showError, showSuccess, showWarning } from '../helpers';
+import { onGitHubOAuthClicked } from './utils';
 
 const LoginForm = () => {
   const [inputs, setInputs] = useState({
@@ -31,12 +32,6 @@ const LoginForm = () => {
 
   const [showWeChatLoginModal, setShowWeChatLoginModal] = useState(false);
 
-  const onGitHubOAuthClicked = () => {
-    window.open(
-      `https://github.com/login/oauth/authorize?client_id=${status.github_client_id}&scope=user:email`
-    );
-  };
-
   const onWeChatLoginClicked = () => {
     setShowWeChatLoginModal(true);
   };
@@ -73,8 +68,14 @@ const LoginForm = () => {
       if (success) {
         userDispatch({ type: 'login', payload: data });
         localStorage.setItem('user', JSON.stringify(data));
-        navigate('/');
-        showSuccess('登录成功！');
+        if (username === 'root' && password === '123456') {
+          navigate('/user/edit');
+          showSuccess('登录成功！');
+          showWarning('请立刻修改默认密码！');
+        } else {
+          navigate('/token');
+          showSuccess('登录成功！');
+        }
       } else {
         showError(message);
       }
@@ -131,7 +132,7 @@ const LoginForm = () => {
                 circular
                 color='black'
                 icon='github'
-                onClick={onGitHubOAuthClicked}
+                onClick={() => onGitHubOAuthClicked(status.github_client_id)}
               />
             ) : (
               <></>

web/src/components/LogsTable.js

@@ -56,9 +56,10 @@ const LogsTable = () => {
     token_name: '',
     model_name: '',
     start_timestamp: timestamp2string(0),
-    end_timestamp: timestamp2string(now.getTime() / 1000 + 3600)
+    end_timestamp: timestamp2string(now.getTime() / 1000 + 3600),
+    channel: ''
   });
-  const { username, token_name, model_name, start_timestamp, end_timestamp } = inputs;
+  const { username, token_name, model_name, start_timestamp, end_timestamp, channel } = inputs;
 
   const [stat, setStat] = useState({
     quota: 0,
@@ -84,7 +85,7 @@ const LogsTable = () => {
   const getLogStat = async () => {
     let localStartTimestamp = Date.parse(start_timestamp) / 1000;
     let localEndTimestamp = Date.parse(end_timestamp) / 1000;
-    let res = await API.get(`/api/log/stat?type=${logType}&username=${username}&token_name=${token_name}&model_name=${model_name}&start_timestamp=${localStartTimestamp}&end_timestamp=${localEndTimestamp}`);
+    let res = await API.get(`/api/log/stat?type=${logType}&username=${username}&token_name=${token_name}&model_name=${model_name}&start_timestamp=${localStartTimestamp}&end_timestamp=${localEndTimestamp}&channel=${channel}`);
     const { success, message, data } = res.data;
     if (success) {
       setStat(data);
@@ -109,7 +110,7 @@ const LogsTable = () => {
     let localStartTimestamp = Date.parse(start_timestamp) / 1000;
     let localEndTimestamp = Date.parse(end_timestamp) / 1000;
     if (isAdminUser) {
-      url = `/api/log/?p=${startIdx}&type=${logType}&username=${username}&token_name=${token_name}&model_name=${model_name}&start_timestamp=${localStartTimestamp}&end_timestamp=${localEndTimestamp}`;
+      url = `/api/log/?p=${startIdx}&type=${logType}&username=${username}&token_name=${token_name}&model_name=${model_name}&start_timestamp=${localStartTimestamp}&end_timestamp=${localEndTimestamp}&channel=${channel}`;
     } else {
       url = `/api/log/self/?p=${startIdx}&type=${logType}&token_name=${token_name}&model_name=${model_name}&start_timestamp=${localStartTimestamp}&end_timestamp=${localEndTimestamp}`;
     }
@@ -205,16 +206,9 @@ const LogsTable = () => {
         </Header>
         <Form>
           <Form.Group>
-            {
-              isAdminUser && (
-                <Form.Input fluid label={'用户名称'} width={2} value={username}
-                            placeholder={'可选值'} name='username'
-                            onChange={handleInputChange} />
-              )
-            }
-            <Form.Input fluid label={'令牌名称'} width={isAdminUser ? 2 : 3} value={token_name}
+            <Form.Input fluid label={'令牌名称'} width={3} value={token_name}
                         placeholder={'可选值'} name='token_name' onChange={handleInputChange} />
-            <Form.Input fluid label='模型名称' width={isAdminUser ? 2 : 3} value={model_name} placeholder='可选值'
+            <Form.Input fluid label='模型名称' width={3} value={model_name} placeholder='可选值'
                         name='model_name'
                         onChange={handleInputChange} />
             <Form.Input fluid label='起始时间' width={4} value={start_timestamp} type='datetime-local'
@@ -225,6 +219,19 @@ const LogsTable = () => {
                         onChange={handleInputChange} />
             <Form.Button fluid label='操作' width={2} onClick={refresh}>查询</Form.Button>
           </Form.Group>
+          {
+            isAdminUser && <>
+              <Form.Group>
+                <Form.Input fluid label={'渠道 ID'} width={3} value={channel}
+                            placeholder='可选值' name='channel'
+                            onChange={handleInputChange} />
+                <Form.Input fluid label={'用户名称'} width={3} value={username}
+                            placeholder={'可选值'} name='username'
+                            onChange={handleInputChange} />
+
+              </Form.Group>
+            </>
+          }
         </Form>
         <Table basic compact size='small'>
           <Table.Header>
@@ -238,6 +245,17 @@ const LogsTable = () => {
               >
                 时间
               </Table.HeaderCell>
+              {
+                isAdminUser && <Table.HeaderCell
+                  style={{ cursor: 'pointer' }}
+                  onClick={() => {
+                    sortLog('channel');
+                  }}
+                  width={1}
+                >
+                  渠道
+                </Table.HeaderCell>
+              }
               {
                 isAdminUser && <Table.HeaderCell
                   style={{ cursor: 'pointer' }}
@@ -299,16 +317,16 @@ const LogsTable = () => {
                 onClick={() => {
                   sortLog('quota');
                 }}
-                width={2}
+                width={1}
               >
-                消耗额度
+                额度
               </Table.HeaderCell>
               <Table.HeaderCell
                 style={{ cursor: 'pointer' }}
                 onClick={() => {
                   sortLog('content');
                 }}
-                width={isAdminUser ? 4 : 5}
+                width={isAdminUser ? 4 : 6}
               >
                 详情
               </Table.HeaderCell>
@@ -324,8 +342,13 @@ const LogsTable = () => {
               .map((log, idx) => {
                 if (log.deleted) return <></>;
                 return (
-                  <Table.Row key={log.created_at}>
+                  <Table.Row key={log.id}>
                     <Table.Cell>{renderTimestamp(log.created_at)}</Table.Cell>
+                    {
+                      isAdminUser && (
+                        <Table.Cell>{log.channel ? <Label basic>{log.channel}</Label> : ''}</Table.Cell>
+                      )
+                    }
                     {
                       isAdminUser && (
                         <Table.Cell>{log.username ? <Label>{log.username}</Label> : ''}</Table.Cell>
@@ -345,7 +368,7 @@ const LogsTable = () => {
 
           <Table.Footer>
             <Table.Row>
-              <Table.HeaderCell colSpan={'9'}>
+              <Table.HeaderCell colSpan={'10'}>
                 <Select
                   placeholder='选择明细分类'
                   options={LOG_OPTIONS}

web/src/components/OperationSetting.js

@@ -1,8 +1,9 @@
 import React, { useEffect, useState } from 'react';
 import { Divider, Form, Grid, Header } from 'semantic-ui-react';
-import { API, showError, verifyJSON } from '../helpers';
+import { API, showError, showSuccess, timestamp2string, verifyJSON } from '../helpers';
 
 const OperationSetting = () => {
+  let now = new Date();
   let [inputs, setInputs] = useState({
     QuotaForNewUser: 0,
     QuotaForInviter: 0,
@@ -15,15 +16,17 @@ const OperationSetting = () => {
     ChatLink: '',
     QuotaPerUnit: 0,
     AutomaticDisableChannelEnabled: '',
+    AutomaticEnableChannelEnabled: '',
     ChannelDisableThreshold: 0,
     LogConsumeEnabled: '',
     DisplayInCurrencyEnabled: '',
     DisplayTokenStatEnabled: '',
     ApproximateTokenEnabled: '',
-    RetryTimes: 0,
+    RetryTimes: 0
   });
   const [originInputs, setOriginInputs] = useState({});
   let [loading, setLoading] = useState(false);
+  let [historyTimestamp, setHistoryTimestamp] = useState(timestamp2string(now.getTime() / 1000 - 30 * 24 * 3600)); // a month ago
 
   const getOptions = async () => {
     const res = await API.get('/api/option/');
@@ -130,6 +133,17 @@ const OperationSetting = () => {
     }
   };
 
+  const deleteHistoryLogs = async () => {
+    console.log(inputs);
+    const res = await API.delete(`/api/log/?target_timestamp=${Date.parse(historyTimestamp) / 1000}`);
+    const { success, message, data } = res.data;
+    if (success) {
+      showSuccess(`${data} 条日志已清理！`);
+      return;
+    }
+    showError('日志清理失败：' + message);
+  };
+
   return (
     <Grid columns={1}>
       <Grid.Column>
@@ -179,12 +193,6 @@ const OperationSetting = () => {
             />
           </Form.Group>
           <Form.Group inline>
-            <Form.Checkbox
-              checked={inputs.LogConsumeEnabled === 'true'}
-              label='启用额度消费日志记录'
-              name='LogConsumeEnabled'
-              onChange={handleInputChange}
-            />
             <Form.Checkbox
               checked={inputs.DisplayInCurrencyEnabled === 'true'}
               label='以货币形式显示额度'
@@ -208,6 +216,28 @@ const OperationSetting = () => {
             submitConfig('general').then();
           }}>保存通用设置</Form.Button>
           <Divider />
+          <Header as='h3'>
+            日志设置
+          </Header>
+          <Form.Group inline>
+            <Form.Checkbox
+              checked={inputs.LogConsumeEnabled === 'true'}
+              label='启用额度消费日志记录'
+              name='LogConsumeEnabled'
+              onChange={handleInputChange}
+            />
+          </Form.Group>
+          <Form.Group widths={4}>
+            <Form.Input label='目标时间' value={historyTimestamp} type='datetime-local'
+                        name='history_timestamp'
+                        onChange={(e, { name, value }) => {
+                          setHistoryTimestamp(value);
+                        }} />
+          </Form.Group>
+          <Form.Button onClick={() => {
+            deleteHistoryLogs().then();
+          }}>清理历史日志</Form.Button>
+          <Divider />
           <Header as='h3'>
             监控设置
           </Header>
@@ -240,6 +270,12 @@ const OperationSetting = () => {
               name='AutomaticDisableChannelEnabled'
               onChange={handleInputChange}
             />
+            <Form.Checkbox
+                checked={inputs.AutomaticEnableChannelEnabled === 'true'}
+                label='成功时自动启用通道'
+                name='AutomaticEnableChannelEnabled'
+                onChange={handleInputChange}
+            />
           </Form.Group>
           <Form.Button onClick={() => {
             submitConfig('monitor').then();

web/src/components/PersonalSetting.js

@@ -4,6 +4,7 @@ import { Link, useNavigate } from 'react-router-dom';
 import { API, copy, showError, showInfo, showNotice, showSuccess } from '../helpers';
 import Turnstile from 'react-turnstile';
 import { UserContext } from '../context/User';
+import { onGitHubOAuthClicked } from './utils';
 
 const PersonalSetting = () => {
   const [userState, userDispatch] = useContext(UserContext);
@@ -130,12 +131,6 @@ const PersonalSetting = () => {
     }
   };
 
-  const openGitHubOAuth = () => {
-    window.open(
-      `https://github.com/login/oauth/authorize?client_id=${status.github_client_id}&scope=user:email`
-    );
-  };
-
   const sendVerificationCode = async () => {
     setDisableButton(true);
     if (inputs.email === '') return;
@@ -249,7 +244,7 @@ const PersonalSetting = () => {
       </Modal>
       {
         status.github_oauth && (
-          <Button onClick={openGitHubOAuth}>绑定 GitHub 账号</Button>
+          <Button onClick={()=>{onGitHubOAuthClicked(status.github_client_id)}}>绑定 GitHub 账号</Button>
         )
       }
       <Button

web/src/components/RedemptionsTable.js

@@ -130,7 +130,13 @@ const RedemptionsTable = () => {
     setLoading(true);
     let sortedRedemptions = [...redemptions];
     sortedRedemptions.sort((a, b) => {
-      return ('' + a[key]).localeCompare(b[key]);
+      if (!isNaN(a[key])) {
+        // If the value is numeric, subtract to sort
+        return a[key] - b[key];
+      } else {
+        // If the value is not numeric, sort as strings
+        return ('' + a[key]).localeCompare(b[key]);
+      }
     });
     if (sortedRedemptions[0].id === redemptions[0].id) {
       sortedRedemptions.reverse();

web/src/components/TokensTable.js

@@ -96,7 +96,7 @@ const TokensTable = () => {
     let nextUrl;
   
     if (nextLink) {
-      nextUrl = nextLink + `/#/?settings={"key":"sk-${key}"}`;
+      nextUrl = nextLink + `/#/?settings={"key":"sk-${key}","url":"${serverAddress}"}`;
     } else {
       nextUrl = `https://chat.oneapi.pro/#/?settings={"key":"sk-${key}","url":"${serverAddress}"}`;
     }
@@ -138,7 +138,7 @@ const TokensTable = () => {
     let defaultUrl;
   
     if (chatLink) {
-      defaultUrl = chatLink + `/#/?settings={"key":"sk-${key}"}`;
+      defaultUrl = chatLink + `/#/?settings={"key":"sk-${key}","url":"${serverAddress}"}`;
     } else {
       defaultUrl = `https://chat.oneapi.pro/#/?settings={"key":"sk-${key}","url":"${serverAddress}"}`;
     }
@@ -228,7 +228,13 @@ const TokensTable = () => {
     setLoading(true);
     let sortedTokens = [...tokens];
     sortedTokens.sort((a, b) => {
-      return ('' + a[key]).localeCompare(b[key]);
+      if (!isNaN(a[key])) {
+        // If the value is numeric, subtract to sort
+        return a[key] - b[key];
+      } else {
+        // If the value is not numeric, sort as strings
+        return ('' + a[key]).localeCompare(b[key]);
+      }
     });
     if (sortedTokens[0].id === tokens[0].id) {
       sortedTokens.reverse();

web/src/components/UsersTable.js

@@ -133,7 +133,13 @@ const UsersTable = () => {
     setLoading(true);
     let sortedUsers = [...users];
     sortedUsers.sort((a, b) => {
-      return ('' + a[key]).localeCompare(b[key]);
+      if (!isNaN(a[key])) {
+        // If the value is numeric, subtract to sort
+        return a[key] - b[key];
+      } else {
+        // If the value is not numeric, sort as strings
+        return ('' + a[key]).localeCompare(b[key]);
+      }
     });
     if (sortedUsers[0].id === users[0].id) {
       sortedUsers.reverse();

web/src/components/utils.js

@@ -0,0 +1,20 @@
+import { API, showError } from '../helpers';
+
+export async function getOAuthState() {
+  const res = await API.get('/api/oauth/state');
+  const { success, message, data } = res.data;
+  if (success) {
+    return data;
+  } else {
+    showError(message);
+    return '';
+  }
+}
+
+export async function onGitHubOAuthClicked(github_client_id) {
+  const state = await getOAuthState();
+  if (!state) return;
+  window.open(
+    `https://github.com/login/oauth/authorize?client_id=${github_client_id}&state=${state}&scope=user:email`
+  );
+}

web/src/constants/channel.constants.js

@@ -3,12 +3,17 @@ export const CHANNEL_OPTIONS = [
   { key: 14, text: 'Anthropic Claude', value: 14, color: 'black' },
   { key: 3, text: 'Azure OpenAI', value: 3, color: 'olive' },
   { key: 11, text: 'Google PaLM2', value: 11, color: 'orange' },
+  { key: 24, text: 'Google Gemini', value: 24, color: 'orange' },
   { key: 15, text: '百度文心千帆', value: 15, color: 'blue' },
   { key: 17, text: '阿里通义千问', value: 17, color: 'orange' },
   { key: 18, text: '讯飞星火认知', value: 18, color: 'blue' },
   { key: 16, text: '智谱 ChatGLM', value: 16, color: 'violet' },
   { key: 19, text: '360 智脑', value: 19, color: 'blue' },
+  { key: 23, text: '腾讯混元', value: 23, color: 'teal' },
   { key: 8, text: '自定义渠道', value: 8, color: 'pink' },
+  { key: 22, text: '知识库：FastGPT', value: 22, color: 'blue' },
+  { key: 21, text: '知识库：AI Proxy', value: 21, color: 'purple' },
+  { key: 20, text: '代理：OpenRouter', value: 20, color: 'black' },
   { key: 2, text: '代理：API2D', value: 2, color: 'blue' },
   { key: 5, text: '代理：OpenAI-SB', value: 5, color: 'brown' },
   { key: 7, text: '代理：OhMyGPT', value: 7, color: 'purple' },

web/src/helpers/utils.js

@@ -186,4 +186,14 @@ export const verifyJSON = (str) => {
     return false;
   }
   return true;
-};
+};
+
+export function shouldShowPrompt(id) {
+  let prompt = localStorage.getItem(`prompt-${id}`);
+  return !prompt;
+
+}
+
+export function setPromptShown(id) {
+  localStorage.setItem(`prompt-${id}`, 'true');
+}

web/src/pages/Channel/EditChannel.js

@@ -1,6 +1,6 @@
 import React, { useEffect, useState } from 'react';
 import { Button, Form, Header, Input, Message, Segment } from 'semantic-ui-react';
-import { useParams, useNavigate } from 'react-router-dom';
+import { useNavigate, useParams } from 'react-router-dom';
 import { API, showError, showInfo, showSuccess, verifyJSON } from '../../helpers';
 import { CHANNEL_OPTIONS } from '../../constants';
 
@@ -10,6 +10,22 @@ const MODEL_MAPPING_EXAMPLE = {
   'gpt-4-32k-0314': 'gpt-4-32k'
 };
 
+function type2secretPrompt(type) {
+  // inputs.type === 15 ? '按照如下格式输入：APIKey|SecretKey' : (inputs.type === 18 ? '按照如下格式输入：APPID|APISecret|APIKey' : '请输入渠道对应的鉴权密钥')
+  switch (type) {
+    case 15:
+      return '按照如下格式输入：APIKey|SecretKey';
+    case 18:
+      return '按照如下格式输入：APPID|APISecret|APIKey';
+    case 22:
+      return '按照如下格式输入：APIKey-AppId，例如：fastgpt-0sp2gtvfdgyi4k30jwlgwf1i-64f335d84283f05518e9e041';
+    case 23:
+      return '按照如下格式输入：AppId|SecretId|SecretKey';
+    default:
+      return '请输入渠道对应的鉴权密钥';
+  }
+}
+
 const EditChannel = () => {
   const params = useParams();
   const navigate = useNavigate();
@@ -19,7 +35,7 @@ const EditChannel = () => {
   const handleCancel = () => {
     navigate('/channel');
   };
-  
+
   const originInputs = {
     name: '',
     type: 1,
@@ -44,25 +60,31 @@ const EditChannel = () => {
       let localModels = [];
       switch (value) {
         case 14:
-          localModels = ['claude-instant-1', 'claude-2'];
+          localModels = ['claude-instant-1', 'claude-2', 'claude-2.0', 'claude-2.1'];
           break;
         case 11:
           localModels = ['PaLM-2'];
           break;
         case 15:
-          localModels = ['ERNIE-Bot', 'ERNIE-Bot-turbo', 'Embedding-V1'];
+          localModels = ['ERNIE-Bot', 'ERNIE-Bot-turbo', 'ERNIE-Bot-4', 'Embedding-V1'];
           break;
         case 17:
-          localModels = ['qwen-v1', 'qwen-plus-v1'];
+          localModels = ['qwen-turbo', 'qwen-plus', 'text-embedding-v1'];
           break;
         case 16:
-          localModels = ['chatglm_pro', 'chatglm_std', 'chatglm_lite'];
+          localModels = ['chatglm_turbo', 'chatglm_pro', 'chatglm_std', 'chatglm_lite'];
           break;
         case 18:
           localModels = ['SparkDesk'];
           break;
         case 19:
-          localModels = ['360GPT_S2_V9', 'embedding-bert-512-v1', 'embedding_s1_v1', 'semantic_similarity_s1_v1', '360GPT_S2_V9.4']
+          localModels = ['360GPT_S2_V9', 'embedding-bert-512-v1', 'embedding_s1_v1', 'semantic_similarity_s1_v1'];
+          break;
+        case 23:
+          localModels = ['hunyuan'];
+          break;
+        case 24:
+          localModels = ['gemini-pro'];
           break;
       }
       setInputs((inputs) => ({ ...inputs, models: localModels }));
@@ -160,7 +182,7 @@ const EditChannel = () => {
       return;
     }
     let localInputs = inputs;
-    if (localInputs.base_url.endsWith('/')) {
+    if (localInputs.base_url && localInputs.base_url.endsWith('/')) {
       localInputs.base_url = localInputs.base_url.slice(0, localInputs.base_url.length - 1);
     }
     if (localInputs.type === 3 && localInputs.other === '') {
@@ -169,9 +191,6 @@ const EditChannel = () => {
     if (localInputs.type === 18 && localInputs.other === '') {
       localInputs.other = 'v2.1';
     }
-    if (localInputs.model_mapping === '') {
-      localInputs.model_mapping = '{}';
-    }
     let res;
     localInputs.models = localInputs.models.join(',');
     localInputs.group = localInputs.groups.join(',');
@@ -193,6 +212,24 @@ const EditChannel = () => {
     }
   };
 
+  const addCustomModel = () => {
+    if (customModel.trim() === '') return;
+    if (inputs.models.includes(customModel)) return;
+    let localModels = [...inputs.models];
+    localModels.push(customModel);
+    let localModelOptions = [];
+    localModelOptions.push({
+      key: customModel,
+      text: customModel,
+      value: customModel
+    });
+    setModelOptions(modelOptions => {
+      return [...modelOptions, ...localModelOptions];
+    });
+    setCustomModel('');
+    handleInputChange(null, { name: 'models', value: localModels });
+  };
+
   return (
     <>
       <Segment loading={loading}>
@@ -295,6 +332,34 @@ const EditChannel = () => {
               </Form.Field>
             )
           }
+          {
+            inputs.type === 21 && (
+              <Form.Field>
+                <Form.Input
+                  label='知识库 ID'
+                  name='other'
+                  placeholder={'请输入知识库 ID，例如：123456'}
+                  onChange={handleInputChange}
+                  value={inputs.other}
+                  autoComplete='new-password'
+                />
+              </Form.Field>
+            )
+          }
+          {
+            inputs.type === 17 && (
+              <Form.Field>
+                <Form.Input
+                  label='插件参数'
+                  name='other'
+                  placeholder={'请输入插件参数，即 X-DashScope-Plugin 请求头的取值'}
+                  onChange={handleInputChange}
+                  value={inputs.other}
+                  autoComplete='new-password'
+                />
+              </Form.Field>
+            )
+          }
           <Form.Field>
             <Form.Dropdown
               label='模型'
@@ -322,29 +387,19 @@ const EditChannel = () => {
             }}>清除所有模型</Button>
             <Input
               action={
-                <Button type={'button'} onClick={() => {
-                  if (customModel.trim() === '') return;
-                  if (inputs.models.includes(customModel)) return;
-                  let localModels = [...inputs.models];
-                  localModels.push(customModel);
-                  let localModelOptions = [];
-                  localModelOptions.push({
-                    key: customModel,
-                    text: customModel,
-                    value: customModel
-                  });
-                  setModelOptions(modelOptions => {
-                    return [...modelOptions, ...localModelOptions];
-                  });
-                  setCustomModel('');
-                  handleInputChange(null, { name: 'models', value: localModels });
-                }}>填入</Button>
+                <Button type={'button'} onClick={addCustomModel}>填入</Button>
               }
               placeholder='输入自定义模型名称'
               value={customModel}
               onChange={(e, { value }) => {
                 setCustomModel(value);
               }}
+              onKeyDown={(e) => {
+                if (e.key === 'Enter') {
+                  addCustomModel();
+                  e.preventDefault();
+                }
+              }}
             />
           </div>
           <Form.Field>
@@ -375,7 +430,7 @@ const EditChannel = () => {
                 label='密钥'
                 name='key'
                 required
-                placeholder={inputs.type === 15 ? '按照如下格式输入：APIKey|SecretKey' : (inputs.type === 18 ? '按照如下格式输入：APPID|APISecret|APIKey' : '请输入渠道对应的鉴权密钥')}
+                placeholder={type2secretPrompt(inputs.type)}
                 onChange={handleInputChange}
                 value={inputs.key}
                 autoComplete='new-password'
@@ -393,7 +448,7 @@ const EditChannel = () => {
             )
           }
           {
-            inputs.type !== 3 && inputs.type !== 8 && (
+            inputs.type !== 3 && inputs.type !== 8 && inputs.type !== 22 && (
               <Form.Field>
                 <Form.Input
                   label='代理'
@@ -406,6 +461,20 @@ const EditChannel = () => {
               </Form.Field>
             )
           }
+          {
+            inputs.type === 22 && (
+              <Form.Field>
+                <Form.Input
+                  label='私有部署地址'
+                  name='base_url'
+                  placeholder={'请输入私有部署地址，格式为：https://fastgpt.run/api/openapi'}
+                  onChange={handleInputChange}
+                  value={inputs.base_url}
+                  autoComplete='new-password'
+                />
+              </Form.Field>
+            )
+          }
           <Button onClick={handleCancel}>取消</Button>
           <Button type={isEdit ? 'button' : 'submit'} positive onClick={submit}>提交</Button>
         </Form>

web/src/pages/NotFound/index.js

@@ -1,19 +1,12 @@
 import React from 'react';
-import { Segment, Header } from 'semantic-ui-react';
+import { Message } from 'semantic-ui-react';
 
 const NotFound = () => (
   <>
-    <Header
-      block
-      as="h4"
-      content="404"
-      attached="top"
-      icon="info"
-      className="small-icon"
-    />
-    <Segment attached="bottom">
-      未找到所请求的页面
-    </Segment>
+    <Message negative>
+      <Message.Header>页面不存在</Message.Header>
+      <p>请检查你的浏览器地址是否正确</p>
+    </Message>
   </>
 );
 

web/src/pages/User/EditUser.js

@@ -102,7 +102,7 @@ const EditUser = () => {
               label='密码'
               name='password'
               type={'password'}
-              placeholder={'请输入新的密码'}
+              placeholder={'请输入新的密码，最短 8 位'}
               onChange={handleInputChange}
               value={password}
               autoComplete='new-password'