Skip to content

phsl / new-api

Go to a project
Commit 9ad40f38 by TAKO Committed by GitHub
Options

Merge branch 'main' into main

parents 210d8886 b869cec7
Showing with 2317 additions and 1551 deletions
.github/workflows/docker-image-amd64.yml
......@@ -18,20 +18,20 @@ jobs:
contents: read
steps:
- name: Check out the repo
uses: actions/checkout@v3
uses: actions/checkout@v4
- name: Save version info
run: |
git describe --tags > VERSION
- name: Log in to Docker Hub
uses: docker/login-action@v2
uses: docker/login-action@v3
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Log in to the Container registry
uses: docker/login-action@v2
uses: docker/login-action@v3
with:
registry: ghcr.io
username: ${{ github.actor }}
......@@ -39,14 +39,14 @@ jobs:
- name: Extract metadata (tags, labels) for Docker
id: meta
uses: docker/metadata-action@v4
uses: docker/metadata-action@v5
with:
images: |
calciumion/new-api
ghcr.io/${{ github.repository }}
- name: Build and push Docker images
uses: docker/build-push-action@v3
uses: docker/build-push-action@v5
with:
context: .
push: true
......
.github/workflows/docker-image-arm64.yml
......@@ -4,7 +4,6 @@ on:
push:
tags:
- '*'
- '!*-alpha*'
workflow_dispatch:
inputs:
name:
......@@ -19,26 +18,26 @@ jobs:
contents: read
steps:
- name: Check out the repo
uses: actions/checkout@v3
uses: actions/checkout@v4
- name: Save version info
run: |
git describe --tags > VERSION
- name: Set up QEMU
uses: docker/setup-qemu-action@v2
uses: docker/setup-qemu-action@v3
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v2
uses: docker/setup-buildx-action@v3
- name: Log in to Docker Hub
uses: docker/login-action@v2
uses: docker/login-action@v3
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Log in to the Container registry
uses: docker/login-action@v2
uses: docker/login-action@v3
with:
registry: ghcr.io
username: ${{ github.actor }}
......@@ -46,14 +45,14 @@ jobs:
- name: Extract metadata (tags, labels) for Docker
id: meta
uses: docker/metadata-action@v4
uses: docker/metadata-action@v5
with:
images: |
calciumion/new-api
ghcr.io/${{ github.repository }}
- name: Build and push Docker images
uses: docker/build-push-action@v3
uses: docker/build-push-action@v5
with:
context: .
platforms: linux/amd64,linux/arm64
......
README.md
......@@ -7,7 +7,6 @@
# New API
🍥新一代大模型网关与AI资产管理系统
<a href="https://trendshift.io/repositories/8227" target="_blank"><img src="https://trendshift.io/api/badge/repositories/8227" alt="Calcium-Ion%2Fnew-api | Trendshift" style="width: 250px; height: 55px;" width="250" height="55"/></a>
......@@ -41,39 +40,40 @@
> - 本项目仅供个人学习使用，不保证稳定性，且不提供任何技术支持。
> - 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。
## 📚 文档
详细文档请访问我们的官方Wiki：[https://docs.newapi.pro/](https://docs.newapi.pro/)
## ✨ 主要特性
1. 🎨 全新的UI界面（部分界面还待更新）
2. 🌍 多语言支持（待完善）
3. 🎨 添加[Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy)接口支持，[对接文档](Midjourney.md)
4. 💰 支持在线充值功能，可在系统设置中设置：
- [x] 易支付
5. 🔍 支持用key查询使用额度：
- 配合项目[neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool)可实现用key查询使用
New API提供了丰富的功能，详细特性请参考[维基百科-特性说明](https://docs.newapi.pro/wiki/features-introduction)：
1. 🎨 全新的UI界面
2. 🌍 多语言支持
3. 🎨 支持[Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy)接口，[对接文档](https://docs.newapi.pro/api/relay/image/midjourney)
4. 💰 支持在线充值功能（易支付）
5. 🔍 支持用key查询使用额度（配合[neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool)）
6. 📑 分页支持选择每页显示数量
7. 🔄 兼容原版One API的数据库，可直接使用原版数据库（one-api.db）
8. 💵 支持模型按次数收费，可在 系统设置-运营设置 中设置
9. ⚖️ 支持渠道**加权随机**
7. 🔄 兼容原版One API的数据库
8. 💵 支持模型按次数收费
9. ⚖️ 支持渠道加权随机
10. 📈 数据看板（控制台）
11. 🔒 可设置令牌能调用的模型
12. 🤖 支持Telegram授权登录：
1. 系统设置-配置登录注册-允许通过Telegram登录
2. 对[@Botfather](https://t.me/botfather)输入指令/setdomain
3. 选择你的bot，然后输入http(s)://你的网站地址/login
4. Telegram Bot 名称是bot username 去掉@后的字符串
13. 🎵 添加 [Suno API](https://github.com/Suno-API/Suno-API)接口支持，[对接文档](Suno.md)
14. 🔄 支持Rerank模型，目前兼容Cohere和Jina，可接入Dify，[对接文档](Rerank.md)
15. ⚡ **[OpenAI Realtime API](https://platform.openai.com/docs/guides/realtime/integration)** - 支持OpenAI的Realtime API，支持Azure渠道
16. 支持使用路由/chat2link 进入聊天界面
17. 🧠 支持通过模型名称后缀设置 reasoning effort：
12. 🤖 支持Telegram授权登录
13. 🎵 支持[Suno API](https://github.com/Suno-API/Suno-API)接口，[接口文档](https://docs.newapi.pro/api/suno-music)
14. 🔄 支持Rerank模型（Cohere和Jina），[接口文档](https://docs.newapi.pro/api/jinaai-rerank)
15. ⚡ 支持OpenAI Realtime API（包括Azure渠道），[接口文档](https://docs.newapi.pro/api/openai-realtime)
16. ⚡ 支持Claude Messages 格式，[接口文档](https://docs.newapi.pro/api/anthropic-chat)
17. 支持使用路由/chat2link进入聊天界面
18. 🧠 支持通过模型名称后缀设置 reasoning effort：
1. OpenAI o系列模型
- 添加后缀 `-high` 设置为 high reasoning effort (例如: `o3-mini-high`)
- 添加后缀 `-medium` 设置为 medium reasoning effort (例如: `o3-mini-medium`)
- 添加后缀 `-low` 设置为 low reasoning effort (例如: `o3-mini-low`)
2. Claude 思考模型
- 添加后缀 `-thinking` 启用思考模式 (例如: `claude-3-7-sonnet-20250219-thinking`)
18. 🔄 思考转内容，支持在 `渠道-编辑-渠道额外设置` 中设置 `thinking_to_content` 选项，默认`false`，开启后会将思考内容`reasoning_content`转换为`<think>`标签拼接到内容中返回。
19. 🔄 模型限流，支持在 `系统设置-速率限制设置` 中设置模型限流，支持设置总请求数限制和成功请求数限制
19. 🔄 思考转内容功能
20. 🔄 模型限流功能
20. 💰 缓存计费支持，开启后可以在缓存命中时按照设定的比例计费：
1. 在 `系统设置-运营设置` 中设置 `提示缓存倍率` 选项
2. 在渠道中设置 `提示缓存倍率`，范围 0-1，例如设置为 0.5 表示缓存命中时按照 50% 计费
......@@ -81,155 +81,116 @@
- [x] OpenAI
- [x] Azure
- [x] DeepSeek
- [ ] Claude
- [x] Claude
## 模型支持
此版本额外支持以下模型：
此版本支持多种模型，详情请参考[接口文档-中继接口](https://docs.newapi.pro/api)：
1. 第三方模型 **gpts** （gpt-4-gizmo-*）
2. [Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy)接口，[对接文档](Midjourney.md)
2. [Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy)接口，[接口文档](https://docs.newapi.pro/api/midjourney-proxy-image)
3. 自定义渠道，支持填入完整调用地址
4. [Suno API](https://github.com/Suno-API/Suno-API) 接口，[对接文档](Suno.md)
5. Rerank模型，目前支持[Cohere](https://cohere.ai/)和[Jina](https://jina.ai/)，[对接文档](Rerank.md)
6. Dify
您可以在渠道中添加自定义模型gpt-4-gizmo-*，此模型并非OpenAI官方模型，而是第三方模型，使用官方key无法调用。
## 比原版One API多出的配置
- `GENERATE_DEFAULT_TOKEN`：是否为新注册用户生成初始令牌，默认为 `false`。
- `STREAMING_TIMEOUT`：设置流式一次回复的超时时间，默认为 60 秒。
- `DIFY_DEBUG`：设置 Dify 渠道是否输出工作流和节点信息到客户端，默认为 `true`。
- `FORCE_STREAM_OPTION`：是否覆盖客户端stream_options参数，请求上游返回流模式usage，默认为 `true`，建议开启，不影响客户端传入stream_options参数返回结果。
- `GET_MEDIA_TOKEN`：是否统计图片token，默认为 `true`，关闭后将不再在本地计算图片token，可能会导致和上游计费不同，此项覆盖 `GET_MEDIA_TOKEN_NOT_STREAM` 选项作用。
- `GET_MEDIA_TOKEN_NOT_STREAM`：是否在非流（`stream=false`）情况下统计图片token，默认为 `true`。
- `UPDATE_TASK`：是否更新异步任务（Midjourney、Suno），默认为 `true`，关闭后将不会更新任务进度。
- `COHERE_SAFETY_SETTING`：Cohere模型[安全设置](https://docs.cohere.com/docs/safety-modes#overview)，可选值为 `NONE`, `CONTEXTUAL`, `STRICT`，默认为 `NONE`。
- `GEMINI_VISION_MAX_IMAGE_NUM`：Gemini模型最大图片数量，默认为 `16`，设置为 `-1` 则不限制。
- `MAX_FILE_DOWNLOAD_MB`: 最大文件下载大小，单位 MB，默认为 `20`。
- `CRYPTO_SECRET`：加密密钥，用于加密数据库内容。
- `AZURE_DEFAULT_API_VERSION`：Azure渠道默认API版本，如果渠道设置中未指定API版本，则使用此版本，默认为 `2024-12-01-preview`
- `NOTIFICATION_LIMIT_DURATION_MINUTE`：通知限制的持续时间（分钟），默认为 `10`。
- `NOTIFY_LIMIT_COUNT`：用户通知在指定持续时间内的最大数量，默认为 `2`。
## 已废弃的环境变量
- ~~`GEMINI_MODEL_MAP`（已废弃）~~：改为到`设置-模型相关设置`中设置
- ~~`GEMINI_SAFETY_SETTING`（已废弃）~~：改为到`设置-模型相关设置`中设置
4. [Suno API](https://github.com/Suno-API/Suno-API)接口，[接口文档](https://docs.newapi.pro/api/suno-music)
5. Rerank模型（[Cohere](https://cohere.ai/)和[Jina](https://jina.ai/)），[接口文档](https://docs.newapi.pro/api/jinaai-rerank)
6. Claude Messages 格式，[接口文档](https://docs.newapi.pro/api/anthropic-chat)
7. Dify
## 环境变量配置
详细配置说明请参考[安装指南-环境变量配置](https://docs.newapi.pro/installation/environment-variables)：
- `GENERATE_DEFAULT_TOKEN`：是否为新注册用户生成初始令牌，默认为 `false`
- `STREAMING_TIMEOUT`：流式回复超时时间，默认60秒
- `DIFY_DEBUG`：Dify渠道是否输出工作流和节点信息，默认 `true`
- `FORCE_STREAM_OPTION`：是否覆盖客户端stream_options参数，默认 `true`
- `GET_MEDIA_TOKEN`：是否统计图片token，默认 `true`
- `GET_MEDIA_TOKEN_NOT_STREAM`：非流情况下是否统计图片token，默认 `true`
- `UPDATE_TASK`：是否更新异步任务（Midjourney、Suno），默认 `true`
- `COHERE_SAFETY_SETTING`：Cohere模型安全设置，可选值为 `NONE`, `CONTEXTUAL`, `STRICT`，默认 `NONE`
- `GEMINI_VISION_MAX_IMAGE_NUM`：Gemini模型最大图片数量，默认 `16`
- `MAX_FILE_DOWNLOAD_MB`: 最大文件下载大小，单位MB，默认 `20`
- `CRYPTO_SECRET`：加密密钥，用于加密数据库内容
- `AZURE_DEFAULT_API_VERSION`：Azure渠道默认API版本，默认 `2024-12-01-preview`
- `NOTIFICATION_LIMIT_DURATION_MINUTE`：通知限制持续时间，默认 `10`分钟
- `NOTIFY_LIMIT_COUNT`：用户通知在指定持续时间内的最大数量，默认 `2`
## 部署
详细部署指南请参考[安装指南-部署方式](https://docs.newapi.pro/installation)：
> [!TIP]
> 最新版Docker镜像：`calciumion/new-api:latest`
> 默认账号root 密码123456
### 多机部署
- 必须设置环境变量 `SESSION_SECRET`，否则会导致多机部署时登录状态不一致。
- 如果公用Redis，必须设置 `CRYPTO_SECRET`，否则会导致多机部署时Redis内容无法获取。
### 多机部署注意事项
- 必须设置环境变量 `SESSION_SECRET`，否则会导致多机部署时登录状态不一致
- 如果公用Redis，必须设置 `CRYPTO_SECRET`，否则会导致多机部署时Redis内容无法获取
### 部署要求
- 本地数据库（默认）：SQLite（Docker 部署默认使用 SQLite，必须挂载 `/data` 目录到宿主机）
- 远程数据库：MySQL 版本 >= 5.7.8，PgSQL 版本 >= 9.6
- 本地数据库（默认）：SQLite（Docker部署必须挂载`/data`目录）
- 远程数据库：MySQL版本 >= 5.7.8，PgSQL版本 >= 9.6
### 使用宝塔面板Docker功能部署
安装宝塔面板 (**9.2.0版本**及以上)，前往 [宝塔面板](https://www.bt.cn/new/download.html) 官网，选择正式版的脚本下载安装
安装后登录宝塔面板，在菜单栏中点击 Docker ，首次进入会提示安装 Docker 服务，点击立即安装，按提示完成安装
安装完成后在应用商店中找到 **New-API** ，点击安装，配置基本选项 即可完成安装
[图文教程](BT.md)
### 部署方式
### 基于 Docker 进行部署
> [!TIP]
> 默认管理员账号root 密码123456
#### 使用宝塔面板Docker功能部署
安装宝塔面板（**9.2.0版本**及以上），在应用商店中找到**New-API**安装即可。
[图文教程](BT.md)
### 使用 Docker Compose 部署（推荐）
#### 使用Docker Compose部署（推荐）
```shell
# 下载项目
git clone https://github.com/Calcium-Ion/new-api.git
cd new-api
# 按需编辑 docker-compose.yml
# nano docker-compose.yml
# vim docker-compose.yml
# 按需编辑docker-compose.yml
# 启动
docker-compose up -d
```
#### 更新版本
```shell
docker-compose pull
docker-compose up -d
```
### 直接使用 Docker 镜像
#### 直接使用Docker镜像
```shell
# 使用 SQLite 的部署命令：
# 使用SQLite
docker run --name new-api -d --restart always -p 3000:3000 -e TZ=Asia/Shanghai -v /home/ubuntu/data/new-api:/data calciumion/new-api:latest
# 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。
# 例如：
# 使用MySQL
docker run --name new-api -d --restart always -p 3000:3000 -e SQL_DSN="root:123456@tcp(localhost:3306)/oneapi" -e TZ=Asia/Shanghai -v /home/ubuntu/data/new-api:/data calciumion/new-api:latest
```
#### 更新版本
```shell
# 拉取最新镜像
docker pull calciumion/new-api:latest
# 停止并删除旧容器
docker stop new-api
docker rm new-api
# 使用相同参数运行新容器
docker run --name new-api -d --restart always -p 3000:3000 -e TZ=Asia/Shanghai -v /home/ubuntu/data/new-api:/data calciumion/new-api:latest
```
## 渠道重试与缓存
渠道重试功能已经实现，可以在`设置->运营设置->通用设置`设置重试次数，**建议开启缓存**功能。
或者使用 Watchtower 自动更新（不推荐，可能会导致数据库不兼容）：
```shell
docker run --rm -v /var/run/docker.sock:/var/run/docker.sock containrrr/watchtower -cR
```
## 渠道重试
渠道重试功能已经实现，可以在`设置->运营设置->通用设置`设置重试次数，**建议开启缓存**功能。
如果开启了重试功能，重试使用下一个优先级，以此类推。
### 缓存设置方法
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`
2. `MEMORY_CACHE_ENABLED`：启用内存缓存（如果设置了`REDIS_CONN_STRING`，则无需手动设置），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`
### 为什么有的时候没有重试
这些错误码不会重试：400，504，524
### 我想让400也重试
在`渠道->编辑`中，将`状态码复写`改为
```json
{
"400": "500"
}
```
可以实现400错误转为500错误，从而重试
1. `REDIS_CONN_STRING`：设置Redis作为缓存
2. `MEMORY_CACHE_ENABLED`：启用内存缓存（设置了Redis则无需手动设置）
## Midjourney接口设置文档
[对接文档](Midjourney.md)
## 接口文档
## Suno接口设置文档
[对接文档](Suno.md)
详细接口文档请参考[接口文档](https://docs.newapi.pro/api)：
## 界面截图
![image](https://github.com/user-attachments/assets/a0dcd349-5df8-4dc8-9acf-ca272b239919)
![image](https://github.com/user-attachments/assets/c7d0f7e1-729c-43e2-ac7c-2cb73b0afc8e)
![image](https://github.com/user-attachments/assets/29f81de5-33fc-4fc5-a5ff-f9b54b653c7c)
![image](https://github.com/user-attachments/assets/4fa53e18-d2c5-477a-9b26-b86e44c71e35)
## 交流群
<img src="https://github.com/user-attachments/assets/9ca0bc82-e057-4230-a28d-9f198fa022e3" width="200">
- [聊天接口（Chat）](https://docs.newapi.pro/api/openai-chat)
- [图像接口（Image）](https://docs.newapi.pro/api/openai-image)
- [Midjourney接口](https://docs.newapi.pro/api/midjourney-proxy-image)
- [音乐接口（Music）](https://docs.newapi.pro/api/relay/music)
- [Suno接口](https://docs.newapi.pro/api/suno-music)
- [重排序接口（Rerank）](https://docs.newapi.pro/api/jinaai-rerank)
- [实时对话接口（Realtime）](https://docs.newapi.pro/api/openai-realtime)
- [Claude聊天接口（messages）](https://docs.newapi.pro/api/anthropic-chat)
## 相关项目
- [One API](https://github.com/songquanpeng/one-api)：原版项目
- [Midjourney-Proxy](https://github.com/novicezk/midjourney-proxy)：Midjourney接口支持
- [chatnio](https://github.com/Deeptrain-Community/chatnio)：下一代 AI 一站式 B/C 端解决方案
- [chatnio](https://github.com/Deeptrain-Community/chatnio)：下一代AI一站式B/C端解决方案
- [neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool)：用key查询使用额度
其他基于New API的项目：
- [new-api-horizon](https://github.com/Calcium-Ion/new-api-horizon)：New API高性能优化版，专注于高并发优化，并支持Claude格式
- [VoAPI](https://github.com/VoAPI/VoAPI)：基于New API的前端美化版本，闭源免费
- [new-api-horizon](https://github.com/Calcium-Ion/new-api-horizon)：New API高性能优化版
- [VoAPI](https://github.com/VoAPI/VoAPI)：基于New API的前端美化版本
## 帮助支持
如有问题，请参考[帮助支持](https://docs.newapi.pro/support)：
- [社区交流](https://docs.newapi.pro/support/community-interaction)
- [反馈问题](https://docs.newapi.pro/support/feedback-issues)
- [常见问题](https://docs.newapi.pro/support/faq)
## 🌟 Star History
......
common/constants.go
......@@ -234,6 +234,7 @@ const (
ChannelTypeMokaAI = 44
ChannelTypeVolcEngine = 45
ChannelTypeBaiduV2 = 46
ChannelTypeXinference = 47
ChannelTypeDummy // this one is only for count, do not add any channel after this
)
......@@ -286,4 +287,5 @@ var ChannelBaseURLs = []string{
"https://api.moka.ai", //44
"https://ark.cn-beijing.volces.com", //45
"https://qianfan.baidubce.com", //46
"", //47
}
common/custom-event.go
......@@ -44,7 +44,7 @@ var fieldReplacer = strings.NewReplacer(
"\r", "\\r")
var dataReplacer = strings.NewReplacer(
"\n", "\ndata:",
"\n", "\n",
"\r", "\\r")
type CustomEvent struct {
......
controller/channel-test.go
......@@ -107,7 +107,7 @@ func testChannel(channel *model.Channel, testModel string) (err error, openAIErr
adaptor.Init(info)
convertedRequest, err := adaptor.ConvertRequest(c, info, request)
convertedRequest, err := adaptor.ConvertOpenAIRequest(c, info, request)
if err != nil {
return err, nil
}
......
controller/relay.go
......@@ -148,6 +148,50 @@ func WssRelay(c *gin.Context) {
}
}
func RelayClaude(c *gin.Context) {
//relayMode := constant.Path2RelayMode(c.Request.URL.Path)
requestId := c.GetString(common.RequestIdKey)
group := c.GetString("group")
originalModel := c.GetString("original_model")
var claudeErr *dto.ClaudeErrorWithStatusCode
for i := 0; i <= common.RetryTimes; i++ {
channel, err := getChannel(c, group, originalModel, i)
if err != nil {
common.LogError(c, err.Error())
claudeErr = service.ClaudeErrorWrapperLocal(err, "get_channel_failed", http.StatusInternalServerError)
break
}
claudeErr = claudeRequest(c, channel)
if claudeErr == nil {
return // 成功处理请求，直接返回
}
openaiErr := service.ClaudeErrorToOpenAIError(claudeErr)
go processChannelError(c, channel.Id, channel.Type, channel.Name, channel.GetAutoBan(), openaiErr)
if !shouldRetry(c, openaiErr, common.RetryTimes-i) {
break
}
}
useChannel := c.GetStringSlice("use_channel")
if len(useChannel) > 1 {
retryLogStr := fmt.Sprintf("重试：%s", strings.Trim(strings.Join(strings.Fields(fmt.Sprint(useChannel)), "->"), "[]"))
common.LogInfo(c, retryLogStr)
}
if claudeErr != nil {
claudeErr.Error.Message = common.MessageWithRequestId(claudeErr.Error.Message, requestId)
c.JSON(claudeErr.StatusCode, gin.H{
"type": "error",
"error": claudeErr.Error,
})
}
}
func relayRequest(c *gin.Context, relayMode int, channel *model.Channel) *dto.OpenAIErrorWithStatusCode {
addUsedChannel(c, channel.Id)
requestBody, _ := common.GetRequestBody(c)
......@@ -162,6 +206,13 @@ func wssRequest(c *gin.Context, ws *websocket.Conn, relayMode int, channel *mode
return relay.WssHelper(c, ws)
}
func claudeRequest(c *gin.Context, channel *model.Channel) *dto.ClaudeErrorWithStatusCode {
addUsedChannel(c, channel.Id)
requestBody, _ := common.GetRequestBody(c)
c.Request.Body = io.NopCloser(bytes.NewBuffer(requestBody))
return relay.ClaudeHelper(c)
}
func addUsedChannel(c *gin.Context, channelId int) {
useChannel := c.GetStringSlice("use_channel")
useChannel = append(useChannel, fmt.Sprintf("%d", channelId))
......
dto/claude.go 0 → 100644
package dto
import "encoding/json"
type ClaudeMetadata struct {
UserId string `json:"user_id"`
}
type ClaudeMediaMessage struct {
Type string `json:"type"`
Text *string `json:"text,omitempty"`
Model string `json:"model,omitempty"`
Source *ClaudeMessageSource `json:"source,omitempty"`
Usage *ClaudeUsage `json:"usage,omitempty"`
StopReason *string `json:"stop_reason,omitempty"`
PartialJson *string `json:"partial_json,omitempty"`
Role string `json:"role,omitempty"`
Thinking string `json:"thinking,omitempty"`
Signature string `json:"signature,omitempty"`
Delta string `json:"delta,omitempty"`
// tool_calls
Id string `json:"id,omitempty"`
Name string `json:"name,omitempty"`
Input any `json:"input,omitempty"`
Content json.RawMessage `json:"content,omitempty"`
ToolUseId string `json:"tool_use_id,omitempty"`
}
func (c *ClaudeMediaMessage) SetText(s string) {
c.Text = &s
}
func (c *ClaudeMediaMessage) GetText() string {
if c.Text == nil {
return ""
}
return *c.Text
}
func (c *ClaudeMediaMessage) IsStringContent() bool {
var content string
return json.Unmarshal(c.Content, &content) == nil
}
func (c *ClaudeMediaMessage) GetStringContent() string {
var content string
if err := json.Unmarshal(c.Content, &content); err == nil {
return content
}
return ""
}
func (c *ClaudeMediaMessage) SetContent(content any) {
jsonContent, _ := json.Marshal(content)
c.Content = jsonContent
}
func (c *ClaudeMediaMessage) ParseMediaContent() []ClaudeMediaMessage {
var mediaContent []ClaudeMediaMessage
if err := json.Unmarshal(c.Content, &mediaContent); err == nil {
return mediaContent
}
return make([]ClaudeMediaMessage, 0)
}
type ClaudeMessageSource struct {
Type string `json:"type"`
MediaType string `json:"media_type"`
Data any `json:"data"`
}
type ClaudeMessage struct {
Role string `json:"role"`
Content any `json:"content"`
}
func (c *ClaudeMessage) IsStringContent() bool {
_, ok := c.Content.(string)
return ok
}
func (c *ClaudeMessage) GetStringContent() string {
if c.IsStringContent() {
return c.Content.(string)
}
return ""
}
func (c *ClaudeMessage) SetStringContent(content string) {
c.Content = content
}
func (c *ClaudeMessage) ParseContent() ([]ClaudeMediaMessage, error) {
// map content to []ClaudeMediaMessage
// parse to json
jsonContent, _ := json.Marshal(c.Content)
var contentList []ClaudeMediaMessage
err := json.Unmarshal(jsonContent, &contentList)
if err != nil {
return make([]ClaudeMediaMessage, 0), err
}
return contentList, nil
}
type Tool struct {
Name string `json:"name"`
Description string `json:"description,omitempty"`
InputSchema map[string]interface{} `json:"input_schema"`
}
type InputSchema struct {
Type string `json:"type"`
Properties any `json:"properties,omitempty"`
Required any `json:"required,omitempty"`
}
type ClaudeRequest struct {
Model string `json:"model"`
Prompt string `json:"prompt,omitempty"`
System any `json:"system,omitempty"`
Messages []ClaudeMessage `json:"messages,omitempty"`
MaxTokens uint `json:"max_tokens,omitempty"`
MaxTokensToSample uint `json:"max_tokens_to_sample,omitempty"`
StopSequences []string `json:"stop_sequences,omitempty"`
Temperature *float64 `json:"temperature,omitempty"`
TopP float64 `json:"top_p,omitempty"`
TopK int `json:"top_k,omitempty"`
//ClaudeMetadata `json:"metadata,omitempty"`
Stream bool `json:"stream,omitempty"`
Tools any `json:"tools,omitempty"`
ToolChoice any `json:"tool_choice,omitempty"`
Thinking *Thinking `json:"thinking,omitempty"`
}
type Thinking struct {
Type string `json:"type"`
BudgetTokens int `json:"budget_tokens"`
}
func (c *ClaudeRequest) IsStringSystem() bool {
_, ok := c.System.(string)
return ok
}
func (c *ClaudeRequest) GetStringSystem() string {
if c.IsStringSystem() {
return c.System.(string)
}
return ""
}
func (c *ClaudeRequest) SetStringSystem(system string) {
c.System = system
}
func (c *ClaudeRequest) ParseSystem() []ClaudeMediaMessage {
// map content to []ClaudeMediaMessage
// parse to json
jsonContent, _ := json.Marshal(c.System)
var contentList []ClaudeMediaMessage
if err := json.Unmarshal(jsonContent, &contentList); err == nil {
return contentList
}
return make([]ClaudeMediaMessage, 0)
}
type ClaudeError struct {
Type string `json:"type"`
Message string `json:"message"`
}
type ClaudeErrorWithStatusCode struct {
Error ClaudeError `json:"error"`
StatusCode int `json:"status_code"`
LocalError bool
}
type ClaudeResponse struct {
Id string `json:"id,omitempty"`
Type string `json:"type"`
Role string `json:"role,omitempty"`
Content []ClaudeMediaMessage `json:"content,omitempty"`
Completion string `json:"completion,omitempty"`
StopReason string `json:"stop_reason,omitempty"`
Model string `json:"model,omitempty"`
Error ClaudeError `json:"error,omitempty"`
Usage *ClaudeUsage `json:"usage,omitempty"`
Index *int `json:"index,omitempty"`
ContentBlock *ClaudeMediaMessage `json:"content_block,omitempty"`
Delta *ClaudeMediaMessage `json:"delta,omitempty"`
Message *ClaudeMediaMessage `json:"message,omitempty"`
}
// set index
func (c *ClaudeResponse) SetIndex(i int) {
c.Index = &i
}
// get index
func (c *ClaudeResponse) GetIndex() int {
if c.Index == nil {
return 0
}
return *c.Index
}
type ClaudeUsage struct {
InputTokens int `json:"input_tokens"`
CacheCreationInputTokens int `json:"cache_creation_input_tokens"`
CacheReadInputTokens int `json:"cache_read_input_tokens"`
OutputTokens int `json:"output_tokens"`
}
dto/openai_response.go
package dto
type TextResponseWithError struct {
Id string `json:"id"`
Object string `json:"object"`
Created int64 `json:"created"`
Choices []OpenAITextResponseChoice `json:"choices"`
Data []OpenAIEmbeddingResponseItem `json:"data"`
Model string `json:"model"`
Usage `json:"usage"`
Error OpenAIError `json:"error"`
}
type SimpleResponse struct {
Usage `json:"usage"`
Error OpenAIError `json:"error"`
......@@ -125,6 +114,20 @@ type ChatCompletionsStreamResponse struct {
Usage *Usage `json:"usage"`
}
func (c *ChatCompletionsStreamResponse) IsToolCall() bool {
if len(c.Choices) == 0 {
return false
}
return len(c.Choices[0].Delta.ToolCalls) > 0
}
func (c *ChatCompletionsStreamResponse) GetFirstToolCall() *ToolCallResponse {
if c.IsToolCall() {
return &c.Choices[0].Delta.ToolCalls[0]
}
return nil
}
func (c *ChatCompletionsStreamResponse) Copy() *ChatCompletionsStreamResponse {
choices := make([]ChatCompletionsStreamResponseChoice, len(c.Choices))
copy(choices, c.Choices)
......
dto/realtime.go
......@@ -44,10 +44,11 @@ type RealtimeUsage struct {
}
type InputTokenDetails struct {
CachedTokens int `json:"cached_tokens"`
TextTokens int `json:"text_tokens"`
AudioTokens int `json:"audio_tokens"`
ImageTokens int `json:"image_tokens"`
CachedTokens int `json:"cached_tokens"`
CachedCreationTokens int
TextTokens int `json:"text_tokens"`
AudioTokens int `json:"audio_tokens"`
ImageTokens int `json:"image_tokens"`
}
type OutputTokenDetails struct {
......
go.mod
......@@ -11,6 +11,7 @@ require (
github.com/aws/aws-sdk-go-v2/credentials v1.17.11
github.com/aws/aws-sdk-go-v2/service/bedrockruntime v1.7.4
github.com/bytedance/gopkg v0.0.0-20220118071334-3db87571198b
github.com/bytedance/sonic v1.11.6
github.com/gin-contrib/cors v1.7.2
github.com/gin-contrib/gzip v0.0.6
github.com/gin-contrib/sessions v0.0.5
......@@ -42,7 +43,6 @@ require (
github.com/aws/aws-sdk-go-v2/internal/configsources v1.3.5 // indirect
github.com/aws/aws-sdk-go-v2/internal/endpoints/v2 v2.6.5 // indirect
github.com/aws/smithy-go v1.20.2 // indirect
github.com/bytedance/sonic v1.11.6 // indirect
github.com/bytedance/sonic/loader v0.1.1 // indirect
github.com/cespare/xxhash/v2 v2.3.0 // indirect
github.com/cloudwego/base64x v0.1.4 // indirect
......
middleware/auth.go
......@@ -174,6 +174,14 @@ func TokenAuth() func(c *gin.Context) {
}
c.Request.Header.Set("Authorization", "Bearer "+key)
}
// 检查path包含/v1/messages
if strings.Contains(c.Request.URL.Path, "/v1/messages") {
// 从x-api-key中获取key
key := c.Request.Header.Get("x-api-key")
if key != "" {
c.Request.Header.Set("Authorization", "Bearer "+key)
}
}
key := c.Request.Header.Get("Authorization")
parts := make([]string, 0)
key = strings.TrimPrefix(key, "Bearer ")
......
model/main.go
package model
import (
"github.com/glebarez/sqlite"
"gorm.io/driver/mysql"
"gorm.io/driver/postgres"
"gorm.io/gorm"
"log"
"one-api/common"
"os"
"strings"
"sync"
"time"
"github.com/glebarez/sqlite"
"gorm.io/driver/mysql"
"gorm.io/driver/postgres"
"gorm.io/gorm"
)
var groupCol string
......@@ -60,7 +61,7 @@ func chooseDB(envName string) (*gorm.DB, error) {
}()
dsn := os.Getenv(envName)
if dsn != "" {
if strings.HasPrefix(dsn, "postgres://") {
if strings.HasPrefix(dsn, "postgres://") || strings.HasPrefix(dsn, "postgresql://") {
// Use PostgreSQL
common.SysLog("using PostgreSQL as database")
common.UsingPostgreSQL = true
......
relay/channel/adapter.go
......@@ -13,7 +13,7 @@ type Adaptor interface {
Init(info *relaycommon.RelayInfo)
GetRequestURL(info *relaycommon.RelayInfo) (string, error)
SetupRequestHeader(c *gin.Context, req *http.Header, info *relaycommon.RelayInfo) error
ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error)
ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error)
ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error)
ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.EmbeddingRequest) (any, error)
ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error)
......@@ -22,6 +22,7 @@ type Adaptor interface {
DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode)
GetModelList() []string
GetChannelName() string
ConvertClaudeRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.ClaudeRequest) (any, error)
}
type TaskAdaptor interface {
......
relay/channel/ali/adaptor.go
......@@ -16,6 +16,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
}
......@@ -44,7 +50,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -87,7 +93,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
}
return
......
relay/channel/ali/image.go
......@@ -26,7 +26,7 @@ func oaiImage2Ali(request dto.ImageRequest) *AliImageRequest {
return &imageRequest
}
func updateTask(info *relaycommon.RelayInfo, taskID string, key string) (*AliResponse, error, []byte) {
func updateTask(info *relaycommon.RelayInfo, taskID string) (*AliResponse, error, []byte) {
url := fmt.Sprintf("%s/api/v1/tasks/%s", info.BaseUrl, taskID)
var aliResponse AliResponse
......@@ -36,7 +36,7 @@ func updateTask(info *relaycommon.RelayInfo, taskID string, key string) (*AliRes
return &aliResponse, err, nil
}
req.Header.Set("Authorization", "Bearer "+key)
req.Header.Set("Authorization", "Bearer "+info.ApiKey)
client := &http.Client{}
resp, err := client.Do(req)
......@@ -58,7 +58,7 @@ func updateTask(info *relaycommon.RelayInfo, taskID string, key string) (*AliRes
return &response, nil, responseBody
}
func asyncTaskWait(info *relaycommon.RelayInfo, taskID string, key string) (*AliResponse, []byte, error) {
func asyncTaskWait(info *relaycommon.RelayInfo, taskID string) (*AliResponse, []byte, error) {
waitSeconds := 3
step := 0
maxStep := 20
......@@ -68,7 +68,7 @@ func asyncTaskWait(info *relaycommon.RelayInfo, taskID string, key string) (*Ali
for {
step++
rsp, err, body := updateTask(info, taskID, key)
rsp, err, body := updateTask(info, taskID)
responseBody = body
if err != nil {
return &taskResponse, responseBody, err
......@@ -125,8 +125,6 @@ func responseAli2OpenAIImage(c *gin.Context, response *AliResponse, info *relayc
}
func aliImageHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
apiKey := c.Request.Header.Get("Authorization")
apiKey = strings.TrimPrefix(apiKey, "Bearer ")
responseFormat := c.GetString("response_format")
var aliTaskResponse AliResponse
......@@ -148,7 +146,7 @@ func aliImageHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rela
return service.OpenAIErrorWrapper(errors.New(aliTaskResponse.Message), "ali_async_task_failed", http.StatusInternalServerError), nil
}
aliResponse, _, err := asyncTaskWait(info, aliTaskResponse.Output.TaskId, apiKey)
aliResponse, _, err := asyncTaskWait(info, aliTaskResponse.Output.TaskId)
if err != nil {
return service.OpenAIErrorWrapper(err, "ali_async_task_wait_failed", http.StatusInternalServerError), nil
}
......
relay/channel/api_request.go
......@@ -7,6 +7,7 @@ import (
"github.com/gorilla/websocket"
"io"
"net/http"
common2 "one-api/common"
"one-api/relay/common"
"one-api/relay/constant"
"one-api/service"
......@@ -31,6 +32,9 @@ func DoApiRequest(a Adaptor, c *gin.Context, info *common.RelayInfo, requestBody
if err != nil {
return nil, fmt.Errorf("get request url failed: %w", err)
}
if common2.DebugEnabled {
println("fullRequestURL:", fullRequestURL)
}
req, err := http.NewRequest(c.Request.Method, fullRequestURL, requestBody)
if err != nil {
return nil, fmt.Errorf("new request failed: %w", err)
......
relay/channel/aws/adaptor.go
......@@ -20,6 +20,10 @@ type Adaptor struct {
RequestMode int
}
func (a *Adaptor) ConvertClaudeRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.ClaudeRequest) (any, error) {
return request, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -43,12 +47,12 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
var claudeReq *claude.ClaudeRequest
var claudeReq *dto.ClaudeRequest
var err error
claudeReq, err = claude.RequestOpenAI2ClaudeMessage(*request)
if err != nil {
......
relay/channel/aws/dto.go
package aws
import (
"one-api/relay/channel/claude"
"one-api/dto"
)
type AwsClaudeRequest struct {
// AnthropicVersion should be "bedrock-2023-05-31"
AnthropicVersion string `json:"anthropic_version"`
System string `json:"system,omitempty"`
Messages []claude.ClaudeMessage `json:"messages"`
MaxTokens uint `json:"max_tokens,omitempty"`
Temperature *float64 `json:"temperature,omitempty"`
TopP float64 `json:"top_p,omitempty"`
TopK int `json:"top_k,omitempty"`
StopSequences []string `json:"stop_sequences,omitempty"`
Tools any `json:"tools,omitempty"`
ToolChoice any `json:"tool_choice,omitempty"`
Thinking *claude.Thinking `json:"thinking,omitempty"`
AnthropicVersion string `json:"anthropic_version"`
System any `json:"system,omitempty"`
Messages []dto.ClaudeMessage `json:"messages"`
MaxTokens uint `json:"max_tokens,omitempty"`
Temperature *float64 `json:"temperature,omitempty"`
TopP float64 `json:"top_p,omitempty"`
TopK int `json:"top_k,omitempty"`
StopSequences []string `json:"stop_sequences,omitempty"`
Tools any `json:"tools,omitempty"`
ToolChoice any `json:"tool_choice,omitempty"`
Thinking *dto.Thinking `json:"thinking,omitempty"`
}
func copyRequest(req *claude.ClaudeRequest) *AwsClaudeRequest {
func copyRequest(req *dto.ClaudeRequest) *AwsClaudeRequest {
return &AwsClaudeRequest{
AnthropicVersion: "bedrock-2023-05-31",
System: req.System,
......
relay/channel/aws/relay-aws.go
......@@ -9,7 +9,7 @@ import (
"io"
"net/http"
"one-api/common"
relaymodel "one-api/dto"
"one-api/dto"
"one-api/relay/channel/claude"
relaycommon "one-api/relay/common"
"one-api/relay/helper"
......@@ -39,10 +39,10 @@ func newAwsClient(c *gin.Context, info *relaycommon.RelayInfo) (*bedrockruntime.
return client, nil
}
func wrapErr(err error) *relaymodel.OpenAIErrorWithStatusCode {
return &relaymodel.OpenAIErrorWithStatusCode{
func wrapErr(err error) *dto.OpenAIErrorWithStatusCode {
return &dto.OpenAIErrorWithStatusCode{
StatusCode: http.StatusInternalServerError,
Error: relaymodel.OpenAIError{
Error: dto.OpenAIError{
Message: fmt.Sprintf("%s", err.Error()),
},
}
......@@ -56,7 +56,7 @@ func awsModelID(requestModel string) (string, error) {
return requestModel, nil
}
func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*relaymodel.OpenAIErrorWithStatusCode, *relaymodel.Usage) {
func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
awsCli, err := newAwsClient(c, info)
if err != nil {
return wrapErr(errors.Wrap(err, "newAwsClient")), nil
......@@ -77,7 +77,7 @@ func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*
if !ok {
return wrapErr(errors.New("request not found")), nil
}
claudeReq := claudeReq_.(*claude.ClaudeRequest)
claudeReq := claudeReq_.(*dto.ClaudeRequest)
awsClaudeReq := copyRequest(claudeReq)
awsReq.Body, err = json.Marshal(awsClaudeReq)
if err != nil {
......@@ -89,14 +89,14 @@ func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*
return wrapErr(errors.Wrap(err, "InvokeModel")), nil
}
claudeResponse := new(claude.ClaudeResponse)
claudeResponse := new(dto.ClaudeResponse)
err = json.Unmarshal(awsResp.Body, claudeResponse)
if err != nil {
return wrapErr(errors.Wrap(err, "unmarshal response")), nil
}
openaiResp := claude.ResponseClaude2OpenAI(requestMode, claudeResponse)
usage := relaymodel.Usage{
usage := dto.Usage{
PromptTokens: claudeResponse.Usage.InputTokens,
CompletionTokens: claudeResponse.Usage.OutputTokens,
TotalTokens: claudeResponse.Usage.InputTokens + claudeResponse.Usage.OutputTokens,
......@@ -107,7 +107,7 @@ func awsHandler(c *gin.Context, info *relaycommon.RelayInfo, requestMode int) (*
return nil, &usage
}
func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo, requestMode int) (*relaymodel.OpenAIErrorWithStatusCode, *relaymodel.Usage) {
func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo, requestMode int) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
awsCli, err := newAwsClient(c, info)
if err != nil {
return wrapErr(errors.Wrap(err, "newAwsClient")), nil
......@@ -128,7 +128,7 @@ func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
if !ok {
return wrapErr(errors.New("request not found")), nil
}
claudeReq := claudeReq_.(*claude.ClaudeRequest)
claudeReq := claudeReq_.(*dto.ClaudeRequest)
awsClaudeReq := copyRequest(claudeReq)
awsReq.Body, err = json.Marshal(awsClaudeReq)
......@@ -144,11 +144,14 @@ func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
defer stream.Close()
c.Writer.Header().Set("Content-Type", "text/event-stream")
var usage relaymodel.Usage
var id string
var model string
claudeInfo := &claude.ClaudeResponseInfo{
ResponseId: fmt.Sprintf("chatcmpl-%s", common.GetUUID()),
Created: common.GetTimestamp(),
Model: info.UpstreamModelName,
ResponseText: strings.Builder{},
Usage: &dto.Usage{},
}
isFirst := true
createdTime := common.GetTimestamp()
c.Stream(func(w io.Writer) bool {
event, ok := <-stream.Events()
if !ok {
......@@ -161,33 +164,19 @@ func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
isFirst = false
info.FirstResponseTime = time.Now()
}
claudeResp := new(claude.ClaudeResponse)
err := json.NewDecoder(bytes.NewReader(v.Value.Bytes)).Decode(claudeResp)
claudeResponse := new(dto.ClaudeResponse)
err := json.NewDecoder(bytes.NewReader(v.Value.Bytes)).Decode(claudeResponse)
if err != nil {
common.SysError("error unmarshalling stream response: " + err.Error())
return false
}
response, claudeUsage := claude.StreamResponseClaude2OpenAI(requestMode, claudeResp)
if claudeUsage != nil {
usage.PromptTokens += claudeUsage.InputTokens
usage.CompletionTokens += claudeUsage.OutputTokens
}
response := claude.StreamResponseClaude2OpenAI(requestMode, claudeResponse)
if response == nil {
if !claude.FormatClaudeResponseInfo(RequestModeMessage, claudeResponse, response, claudeInfo) {
return true
}
if response.Id != "" {
id = response.Id
}
if response.Model != "" {
model = response.Model
}
response.Created = createdTime
response.Id = id
response.Model = model
jsonStr, err := json.Marshal(response)
if err != nil {
common.SysError("error marshalling stream response: " + err.Error())
......@@ -203,8 +192,16 @@ func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
return false
}
})
if claudeInfo.Usage.PromptTokens == 0 {
//上游出错
}
if claudeInfo.Usage.CompletionTokens == 0 {
claudeInfo.Usage, _ = service.ResponseText2Usage(claudeInfo.ResponseText.String(), info.UpstreamModelName, claudeInfo.Usage.PromptTokens)
}
if info.ShouldIncludeUsage {
response := helper.GenerateFinalUsageResponse(id, createdTime, info.UpstreamModelName, usage)
response := helper.GenerateFinalUsageResponse(claudeInfo.ResponseId, claudeInfo.Created, info.UpstreamModelName, *claudeInfo.Usage)
err := helper.ObjectData(c, response)
if err != nil {
common.SysError("send final response failed: " + err.Error())
......@@ -217,5 +214,5 @@ func awsStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
return service.OpenAIErrorWrapperLocal(err, "close_response_body_failed", http.StatusInternalServerError), nil
}
}
return nil, &usage
return nil, claudeInfo.Usage
}
relay/channel/baidu/adaptor.go
......@@ -16,6 +16,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -104,7 +110,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......
relay/channel/baidu_v2/adaptor.go
......@@ -15,6 +15,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -38,7 +44,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -62,7 +68,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/channel/claude/adaptor.go
......@@ -22,6 +22,10 @@ type Adaptor struct {
RequestMode int
}
func (a *Adaptor) ConvertClaudeRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.ClaudeRequest) (any, error) {
return request, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -60,7 +64,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......
relay/channel/claude/dto.go
package claude
type ClaudeMetadata struct {
UserId string `json:"user_id"`
}
type ClaudeMediaMessage struct {
Type string `json:"type"`
Text string `json:"text,omitempty"`
Source *ClaudeMessageSource `json:"source,omitempty"`
Usage *ClaudeUsage `json:"usage,omitempty"`
StopReason *string `json:"stop_reason,omitempty"`
PartialJson string `json:"partial_json,omitempty"`
Thinking string `json:"thinking,omitempty"`
Signature string `json:"signature,omitempty"`
Delta string `json:"delta,omitempty"`
// tool_calls
Id string `json:"id,omitempty"`
Name string `json:"name,omitempty"`
Input any `json:"input,omitempty"`
Content string `json:"content,omitempty"`
ToolUseId string `json:"tool_use_id,omitempty"`
}
type ClaudeMessageSource struct {
Type string `json:"type"`
MediaType string `json:"media_type"`
Data string `json:"data"`
}
type ClaudeMessage struct {
Role string `json:"role"`
Content any `json:"content"`
}
type Tool struct {
Name string `json:"name"`
Description string `json:"description,omitempty"`
InputSchema map[string]interface{} `json:"input_schema"`
}
type InputSchema struct {
Type string `json:"type"`
Properties any `json:"properties,omitempty"`
Required any `json:"required,omitempty"`
}
type ClaudeRequest struct {
Model string `json:"model"`
Prompt string `json:"prompt,omitempty"`
System string `json:"system,omitempty"`
Messages []ClaudeMessage `json:"messages,omitempty"`
MaxTokens uint `json:"max_tokens,omitempty"`
MaxTokensToSample uint `json:"max_tokens_to_sample,omitempty"`
StopSequences []string `json:"stop_sequences,omitempty"`
Temperature *float64 `json:"temperature,omitempty"`
TopP float64 `json:"top_p,omitempty"`
TopK int `json:"top_k,omitempty"`
//ClaudeMetadata `json:"metadata,omitempty"`
Stream bool `json:"stream,omitempty"`
Tools any `json:"tools,omitempty"`
ToolChoice any `json:"tool_choice,omitempty"`
Thinking *Thinking `json:"thinking,omitempty"`
}
type Thinking struct {
Type string `json:"type"`
BudgetTokens int `json:"budget_tokens"`
}
type ClaudeError struct {
Type string `json:"type"`
Message string `json:"message"`
}
type ClaudeResponse struct {
Id string `json:"id"`
Type string `json:"type"`
Content []ClaudeMediaMessage `json:"content"`
Completion string `json:"completion"`
StopReason string `json:"stop_reason"`
Model string `json:"model"`
Error ClaudeError `json:"error"`
Usage ClaudeUsage `json:"usage"`
Index int `json:"index"` // stream only
ContentBlock *ClaudeMediaMessage `json:"content_block"`
Delta *ClaudeMediaMessage `json:"delta"` // stream only
Message *ClaudeResponse `json:"message"` // stream only: message_start
}
type ClaudeUsage struct {
InputTokens int `json:"input_tokens"`
OutputTokens int `json:"output_tokens"`
}
//
//type ClaudeMetadata struct {
// UserId string `json:"user_id"`
//}
//
//type ClaudeMediaMessage struct {
// Type string `json:"type"`
// Text string `json:"text,omitempty"`
// Source *ClaudeMessageSource `json:"source,omitempty"`
// Usage *ClaudeUsage `json:"usage,omitempty"`
// StopReason *string `json:"stop_reason,omitempty"`
// PartialJson string `json:"partial_json,omitempty"`
// Thinking string `json:"thinking,omitempty"`
// Signature string `json:"signature,omitempty"`
// Delta string `json:"delta,omitempty"`
// // tool_calls
// Id string `json:"id,omitempty"`
// Name string `json:"name,omitempty"`
// Input any `json:"input,omitempty"`
// Content string `json:"content,omitempty"`
// ToolUseId string `json:"tool_use_id,omitempty"`
//}
//
//type ClaudeMessageSource struct {
// Type string `json:"type"`
// MediaType string `json:"media_type"`
// Data string `json:"data"`
//}
//
//type ClaudeMessage struct {
// Role string `json:"role"`
// Content any `json:"content"`
//}
//
//type Tool struct {
// Name string `json:"name"`
// Description string `json:"description,omitempty"`
// InputSchema map[string]interface{} `json:"input_schema"`
//}
//
//type InputSchema struct {
// Type string `json:"type"`
// Properties any `json:"properties,omitempty"`
// Required any `json:"required,omitempty"`
//}
//
//type ClaudeRequest struct {
// Model string `json:"model"`
// Prompt string `json:"prompt,omitempty"`
// System string `json:"system,omitempty"`
// Messages []ClaudeMessage `json:"messages,omitempty"`
// MaxTokens uint `json:"max_tokens,omitempty"`
// MaxTokensToSample uint `json:"max_tokens_to_sample,omitempty"`
// StopSequences []string `json:"stop_sequences,omitempty"`
// Temperature *float64 `json:"temperature,omitempty"`
// TopP float64 `json:"top_p,omitempty"`
// TopK int `json:"top_k,omitempty"`
// //ClaudeMetadata `json:"metadata,omitempty"`
// Stream bool `json:"stream,omitempty"`
// Tools any `json:"tools,omitempty"`
// ToolChoice any `json:"tool_choice,omitempty"`
// Thinking *Thinking `json:"thinking,omitempty"`
//}
//
//type Thinking struct {
// Type string `json:"type"`
// BudgetTokens int `json:"budget_tokens"`
//}
//
//type ClaudeError struct {
// Type string `json:"type"`
// Message string `json:"message"`
//}
//
//type ClaudeResponse struct {
// Id string `json:"id"`
// Type string `json:"type"`
// Content []ClaudeMediaMessage `json:"content"`
// Completion string `json:"completion"`
// StopReason string `json:"stop_reason"`
// Model string `json:"model"`
// Error ClaudeError `json:"error"`
// Usage ClaudeUsage `json:"usage"`
// Index int `json:"index"` // stream only
// ContentBlock *ClaudeMediaMessage `json:"content_block"`
// Delta *ClaudeMediaMessage `json:"delta"` // stream only
// Message *ClaudeResponse `json:"message"` // stream only: message_start
//}
//
//type ClaudeUsage struct {
// InputTokens int `json:"input_tokens"`
// OutputTokens int `json:"output_tokens"`
//}
relay/channel/claude/relay-claude.go
package claude
import (
"bytes"
"encoding/json"
"fmt"
"io"
......@@ -29,9 +30,9 @@ func stopReasonClaude2OpenAI(reason string) string {
}
}
func RequestOpenAI2ClaudeComplete(textRequest dto.GeneralOpenAIRequest) *ClaudeRequest {
func RequestOpenAI2ClaudeComplete(textRequest dto.GeneralOpenAIRequest) *dto.ClaudeRequest {
claudeRequest := ClaudeRequest{
claudeRequest := dto.ClaudeRequest{
Model: textRequest.Model,
Prompt: "",
StopSequences: nil,
......@@ -60,12 +61,12 @@ func RequestOpenAI2ClaudeComplete(textRequest dto.GeneralOpenAIRequest) *ClaudeR
return &claudeRequest
}
func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*ClaudeRequest, error) {
claudeTools := make([]Tool, 0, len(textRequest.Tools))
func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*dto.ClaudeRequest, error) {
claudeTools := make([]dto.Tool, 0, len(textRequest.Tools))
for _, tool := range textRequest.Tools {
if params, ok := tool.Function.Parameters.(map[string]any); ok {
claudeTool := Tool{
claudeTool := dto.Tool{
Name: tool.Function.Name,
Description: tool.Function.Description,
}
......@@ -83,7 +84,7 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*ClaudeR
}
}
claudeRequest := ClaudeRequest{
claudeRequest := dto.ClaudeRequest{
Model: textRequest.Model,
MaxTokens: textRequest.MaxTokens,
StopSequences: nil,
......@@ -107,7 +108,7 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*ClaudeR
}
// BudgetTokens 为 max_tokens 的 80%
claudeRequest.Thinking = &Thinking{
claudeRequest.Thinking = &dto.Thinking{
Type: "enabled",
BudgetTokens: int(float64(claudeRequest.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage),
}
......@@ -165,7 +166,7 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*ClaudeR
lastMessage = fmtMessage
}
claudeMessages := make([]ClaudeMessage, 0)
claudeMessages := make([]dto.ClaudeMessage, 0)
isFirstMessage := true
for _, message := range formatMessages {
if message.Role == "system" {
......@@ -186,63 +187,63 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*ClaudeR
isFirstMessage = false
if message.Role != "user" {
// fix: first message is assistant, add user message
claudeMessage := ClaudeMessage{
claudeMessage := dto.ClaudeMessage{
Role: "user",
Content: []ClaudeMediaMessage{
Content: []dto.ClaudeMediaMessage{
{
Type: "text",
Text: "...",
Text: common.GetPointer[string]("..."),
},
},
}
claudeMessages = append(claudeMessages, claudeMessage)
}
}
claudeMessage := ClaudeMessage{
claudeMessage := dto.ClaudeMessage{
Role: message.Role,
}
if message.Role == "tool" {
if len(claudeMessages) > 0 && claudeMessages[len(claudeMessages)-1].Role == "user" {
lastMessage := claudeMessages[len(claudeMessages)-1]
if content, ok := lastMessage.Content.(string); ok {
lastMessage.Content = []ClaudeMediaMessage{
lastMessage.Content = []dto.ClaudeMediaMessage{
{
Type: "text",
Text: content,
Text: common.GetPointer[string](content),
},
}
}
lastMessage.Content = append(lastMessage.Content.([]ClaudeMediaMessage), ClaudeMediaMessage{
lastMessage.Content = append(lastMessage.Content.([]dto.ClaudeMediaMessage), dto.ClaudeMediaMessage{
Type: "tool_result",
ToolUseId: message.ToolCallId,
Content: message.StringContent(),
Content: message.Content,
})
claudeMessages[len(claudeMessages)-1] = lastMessage
continue
} else {
claudeMessage.Role = "user"
claudeMessage.Content = []ClaudeMediaMessage{
claudeMessage.Content = []dto.ClaudeMediaMessage{
{
Type: "tool_result",
ToolUseId: message.ToolCallId,
Content: message.StringContent(),
Content: message.Content,
},
}
}
} else if message.IsStringContent() && message.ToolCalls == nil {
claudeMessage.Content = message.StringContent()
} else {
claudeMediaMessages := make([]ClaudeMediaMessage, 0)
claudeMediaMessages := make([]dto.ClaudeMediaMessage, 0)
for _, mediaMessage := range message.ParseContent() {
claudeMediaMessage := ClaudeMediaMessage{
claudeMediaMessage := dto.ClaudeMediaMessage{
Type: mediaMessage.Type,
}
if mediaMessage.Type == "text" {
claudeMediaMessage.Text = mediaMessage.Text
claudeMediaMessage.Text = common.GetPointer[string](mediaMessage.Text)
} else {
imageUrl := mediaMessage.ImageUrl.(dto.MessageImageUrl)
claudeMediaMessage.Type = "image"
claudeMediaMessage.Source = &ClaudeMessageSource{
claudeMediaMessage.Source = &dto.ClaudeMessageSource{
Type: "base64",
}
// 判断是否是url
......@@ -272,7 +273,7 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*ClaudeR
common.SysError("tool call function arguments is not a map[string]any: " + fmt.Sprintf("%v", toolCall.Function.Arguments))
continue
}
claudeMediaMessages = append(claudeMediaMessages, ClaudeMediaMessage{
claudeMediaMessages = append(claudeMediaMessages, dto.ClaudeMediaMessage{
Type: "tool_use",
Id: toolCall.ID,
Name: toolCall.Function.Name,
......@@ -290,9 +291,8 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*ClaudeR
return &claudeRequest, nil
}
func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) (*dto.ChatCompletionsStreamResponse, *ClaudeUsage) {
func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *dto.ClaudeResponse) *dto.ChatCompletionsStreamResponse {
var response dto.ChatCompletionsStreamResponse
var claudeUsage *ClaudeUsage
response.Object = "chat.completion.chunk"
response.Model = claudeResponse.Model
response.Choices = make([]dto.ChatCompletionsStreamResponseChoice, 0)
......@@ -308,7 +308,7 @@ func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) (*
if claudeResponse.Type == "message_start" {
response.Id = claudeResponse.Message.Id
response.Model = claudeResponse.Message.Model
claudeUsage = &claudeResponse.Message.Usage
//claudeUsage = &claudeResponse.Message.Usage
choice.Delta.SetContentString("")
choice.Delta.Role = "assistant"
} else if claudeResponse.Type == "content_block_start" {
......@@ -325,17 +325,17 @@ func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) (*
})
}
} else {
return nil, nil
return nil
}
} else if claudeResponse.Type == "content_block_delta" {
if claudeResponse.Delta != nil {
choice.Index = claudeResponse.Index
choice.Delta.SetContentString(claudeResponse.Delta.Text)
choice.Index = *claudeResponse.Index
choice.Delta.Content = claudeResponse.Delta.Text
switch claudeResponse.Delta.Type {
case "input_json_delta":
tools = append(tools, dto.ToolCallResponse{
Function: dto.FunctionResponse{
Arguments: claudeResponse.Delta.PartialJson,
Arguments: *claudeResponse.Delta.PartialJson,
},
})
case "signature_delta":
......@@ -352,26 +352,23 @@ func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) (*
if finishReason != "null" {
choice.FinishReason = &finishReason
}
claudeUsage = &claudeResponse.Usage
//claudeUsage = &claudeResponse.Usage
} else if claudeResponse.Type == "message_stop" {
return nil, nil
return nil
} else {
return nil, nil
return nil
}
}
if claudeUsage == nil {
claudeUsage = &ClaudeUsage{}
}
if len(tools) > 0 {
choice.Delta.Content = nil // compatible with other OpenAI derivative applications, like LobeOpenAICompatibleFactory ...
choice.Delta.ToolCalls = tools
}
response.Choices = append(response.Choices, choice)
return &response, claudeUsage
return &response
}
func ResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) *dto.OpenAITextResponse {
func ResponseClaude2OpenAI(reqMode int, claudeResponse *dto.ClaudeResponse) *dto.OpenAITextResponse {
choices := make([]dto.OpenAITextResponseChoice, 0)
fullTextResponse := dto.OpenAITextResponse{
Id: fmt.Sprintf("chatcmpl-%s", common.GetUUID()),
......@@ -379,8 +376,10 @@ func ResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) *dto.Ope
Created: common.GetTimestamp(),
}
var responseText string
var responseThinking string
if len(claudeResponse.Content) > 0 {
responseText = claudeResponse.Content[0].Text
responseText = claudeResponse.Content[0].GetText()
responseThinking = claudeResponse.Content[0].Thinking
}
tools := make([]dto.ToolCallResponse, 0)
thinkingContent := ""
......@@ -415,7 +414,7 @@ func ResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) *dto.Ope
// 加密的不管， 只输出明文的推理过程
thinkingContent = message.Thinking
case "text":
responseText = message.Text
responseText = *message.Text
}
}
}
......@@ -427,6 +426,9 @@ func ResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) *dto.Ope
FinishReason: stopReasonClaude2OpenAI(claudeResponse.StopReason),
}
choice.SetStringContent(responseText)
if len(responseThinking) > 0 {
choice.ReasoningContent = responseThinking
}
if len(tools) > 0 {
choice.Message.SetToolCalls(tools)
}
......@@ -437,49 +439,124 @@ func ResponseClaude2OpenAI(reqMode int, claudeResponse *ClaudeResponse) *dto.Ope
return &fullTextResponse
}
type ClaudeResponseInfo struct {
ResponseId string
Created int64
Model string
ResponseText strings.Builder
Usage *dto.Usage
}
func FormatClaudeResponseInfo(requestMode int, claudeResponse *dto.ClaudeResponse, oaiResponse *dto.ChatCompletionsStreamResponse, claudeInfo *ClaudeResponseInfo) bool {
if requestMode == RequestModeCompletion {
claudeInfo.ResponseText.WriteString(claudeResponse.Completion)
} else {
if claudeResponse.Type == "message_start" {
// message_start, 获取usage
claudeInfo.ResponseId = claudeResponse.Message.Id
claudeInfo.Model = claudeResponse.Message.Model
claudeInfo.Usage.PromptTokens = claudeResponse.Message.Usage.InputTokens
} else if claudeResponse.Type == "content_block_delta" {
if claudeResponse.Delta.Text != nil {
claudeInfo.ResponseText.WriteString(*claudeResponse.Delta.Text)
}
} else if claudeResponse.Type == "message_delta" {
claudeInfo.Usage.CompletionTokens = claudeResponse.Usage.OutputTokens
if claudeResponse.Usage.InputTokens > 0 {
claudeInfo.Usage.PromptTokens = claudeResponse.Usage.InputTokens
}
claudeInfo.Usage.TotalTokens = claudeInfo.Usage.PromptTokens + claudeResponse.Usage.OutputTokens
} else if claudeResponse.Type == "content_block_start" {
} else {
return false
}
}
if oaiResponse != nil {
oaiResponse.Id = claudeInfo.ResponseId
oaiResponse.Created = claudeInfo.Created
oaiResponse.Model = claudeInfo.Model
}
return true
}
func ClaudeStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo, requestMode int) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
responseId := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
var usage *dto.Usage
usage = &dto.Usage{}
responseText := ""
createdTime := common.GetTimestamp()
if info.RelayFormat == relaycommon.RelayFormatOpenAI {
return toOpenAIStreamHandler(c, resp, info, requestMode)
}
usage := &dto.Usage{}
responseText := strings.Builder{}
helper.StreamScannerHandler(c, resp, info, func(data string) bool {
var claudeResponse ClaudeResponse
err := json.Unmarshal([]byte(data), &claudeResponse)
var claudeResponse dto.ClaudeResponse
err := json.NewDecoder(bytes.NewReader(common.StringToByteSlice(data))).Decode(&claudeResponse)
if err != nil {
common.SysError("error unmarshalling stream response: " + err.Error())
return true
}
response, claudeUsage := StreamResponseClaude2OpenAI(requestMode, &claudeResponse)
if response == nil {
return true
}
if requestMode == RequestModeCompletion {
responseText += claudeResponse.Completion
responseId = response.Id
responseText.WriteString(claudeResponse.Completion)
} else {
if claudeResponse.Type == "message_start" {
// message_start, 获取usage
responseId = claudeResponse.Message.Id
info.UpstreamModelName = claudeResponse.Message.Model
usage.PromptTokens = claudeUsage.InputTokens
usage.PromptTokens = claudeResponse.Message.Usage.InputTokens
usage.PromptTokensDetails.CachedTokens = claudeResponse.Message.Usage.CacheReadInputTokens
usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Message.Usage.CacheCreationInputTokens
usage.CompletionTokens = claudeResponse.Message.Usage.OutputTokens
} else if claudeResponse.Type == "content_block_delta" {
responseText += claudeResponse.Delta.Text
responseText.WriteString(claudeResponse.Delta.GetText())
} else if claudeResponse.Type == "message_delta" {
usage.CompletionTokens = claudeUsage.OutputTokens
usage.TotalTokens = claudeUsage.InputTokens + claudeUsage.OutputTokens
} else if claudeResponse.Type == "content_block_start" {
return true
} else {
return true
if claudeResponse.Usage.InputTokens > 0 {
// 不叠加，只取最新的
usage.PromptTokens = claudeResponse.Usage.InputTokens
}
usage.CompletionTokens = claudeResponse.Usage.OutputTokens
usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
}
}
//response.Id = responseId
response.Id = responseId
response.Created = createdTime
response.Model = info.UpstreamModelName
helper.ClaudeChunkData(c, claudeResponse, data)
return true
})
if requestMode == RequestModeCompletion {
usage, _ = service.ResponseText2Usage(responseText.String(), info.UpstreamModelName, info.PromptTokens)
} else {
// 说明流模式建立失败，可能为官方出错
if usage.PromptTokens == 0 {
//usage.PromptTokens = info.PromptTokens
}
if usage.CompletionTokens == 0 {
usage, _ = service.ResponseText2Usage(responseText.String(), info.UpstreamModelName, usage.PromptTokens)
}
}
return nil, usage
}
func toOpenAIStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo, requestMode int) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
responseId := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
claudeInfo := &ClaudeResponseInfo{
ResponseId: responseId,
Created: common.GetTimestamp(),
Model: info.UpstreamModelName,
ResponseText: strings.Builder{},
Usage: &dto.Usage{},
}
helper.StreamScannerHandler(c, resp, info, func(data string) bool {
var claudeResponse dto.ClaudeResponse
err := json.NewDecoder(bytes.NewReader(common.StringToByteSlice(data))).Decode(&claudeResponse)
if err != nil {
common.SysError("error unmarshalling stream response: " + err.Error())
return true
}
response := StreamResponseClaude2OpenAI(requestMode, &claudeResponse)
if !FormatClaudeResponseInfo(requestMode, &claudeResponse, response, claudeInfo) {
return true
}
err = helper.ObjectData(c, response)
if err != nil {
......@@ -489,25 +566,24 @@ func ClaudeStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.
})
if requestMode == RequestModeCompletion {
usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, info.PromptTokens)
claudeInfo.Usage, _ = service.ResponseText2Usage(claudeInfo.ResponseText.String(), info.UpstreamModelName, info.PromptTokens)
} else {
if usage.PromptTokens == 0 {
usage.PromptTokens = info.PromptTokens
if claudeInfo.Usage.PromptTokens == 0 {
//上游出错
}
if usage.CompletionTokens == 0 {
usage, _ = service.ResponseText2Usage(responseText, info.UpstreamModelName, usage.PromptTokens)
if claudeInfo.Usage.CompletionTokens == 0 {
claudeInfo.Usage, _ = service.ResponseText2Usage(claudeInfo.ResponseText.String(), info.UpstreamModelName, claudeInfo.Usage.PromptTokens)
}
}
if info.ShouldIncludeUsage {
response := helper.GenerateFinalUsageResponse(responseId, createdTime, info.UpstreamModelName, *usage)
response := helper.GenerateFinalUsageResponse(responseId, claudeInfo.Created, info.UpstreamModelName, *claudeInfo.Usage)
err := helper.ObjectData(c, response)
if err != nil {
common.SysError("send final response failed: " + err.Error())
}
}
helper.Done(c)
//resp.Body.Close()
return nil, usage
return nil, claudeInfo.Usage
}
func ClaudeHandler(c *gin.Context, resp *http.Response, requestMode int, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
......@@ -519,7 +595,10 @@ func ClaudeHandler(c *gin.Context, resp *http.Response, requestMode int, info *r
if err != nil {
return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
}
var claudeResponse ClaudeResponse
if common.DebugEnabled {
println("responseBody: ", string(responseBody))
}
var claudeResponse dto.ClaudeResponse
err = json.Unmarshal(responseBody, &claudeResponse)
if err != nil {
return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
......@@ -535,13 +614,12 @@ func ClaudeHandler(c *gin.Context, resp *http.Response, requestMode int, info *r
StatusCode: resp.StatusCode,
}, nil
}
fullTextResponse := ResponseClaude2OpenAI(requestMode, &claudeResponse)
completionTokens, err := service.CountTextToken(claudeResponse.Completion, info.OriginModelName)
if err != nil {
return service.OpenAIErrorWrapper(err, "count_token_text_failed", http.StatusInternalServerError), nil
}
usage := dto.Usage{}
if requestMode == RequestModeCompletion {
completionTokens, err := service.CountTextToken(claudeResponse.Completion, info.OriginModelName)
if err != nil {
return service.OpenAIErrorWrapper(err, "count_token_text_failed", http.StatusInternalServerError), nil
}
usage.PromptTokens = info.PromptTokens
usage.CompletionTokens = completionTokens
usage.TotalTokens = info.PromptTokens + completionTokens
......@@ -549,14 +627,23 @@ func ClaudeHandler(c *gin.Context, resp *http.Response, requestMode int, info *r
usage.PromptTokens = claudeResponse.Usage.InputTokens
usage.CompletionTokens = claudeResponse.Usage.OutputTokens
usage.TotalTokens = claudeResponse.Usage.InputTokens + claudeResponse.Usage.OutputTokens
usage.PromptTokensDetails.CachedTokens = claudeResponse.Usage.CacheReadInputTokens
usage.PromptTokensDetails.CachedCreationTokens = claudeResponse.Usage.CacheCreationInputTokens
}
fullTextResponse.Usage = usage
jsonResponse, err := json.Marshal(fullTextResponse)
if err != nil {
return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
var responseData []byte
switch info.RelayFormat {
case relaycommon.RelayFormatOpenAI:
openaiResponse := ResponseClaude2OpenAI(requestMode, &claudeResponse)
openaiResponse.Usage = usage
responseData, err = json.Marshal(openaiResponse)
if err != nil {
return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
}
case relaycommon.RelayFormatClaude:
responseData = responseBody
}
c.Writer.Header().Set("Content-Type", "application/json")
c.Writer.WriteHeader(resp.StatusCode)
_, err = c.Writer.Write(jsonResponse)
_, err = c.Writer.Write(responseData)
return nil, &usage
}
relay/channel/cloudflare/adaptor.go
......@@ -17,6 +17,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
}
......@@ -37,7 +43,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......
relay/channel/cohere/adaptor.go
......@@ -15,6 +15,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -42,7 +48,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
return requestOpenAI2Cohere(*request), nil
}
......@@ -59,7 +65,6 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
if info.RelayMode == constant.RelayModeRerank {
err, usage = cohereRerankHandler(c, resp, info)
......
relay/channel/deepseek/adaptor.go
......@@ -16,6 +16,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -44,7 +50,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -68,7 +74,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/channel/dify/adaptor.go
......@@ -9,7 +9,6 @@ import (
"one-api/dto"
"one-api/relay/channel"
relaycommon "one-api/relay/common"
"strings"
)
const (
......@@ -23,6 +22,12 @@ type Adaptor struct {
BotType int
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -34,15 +39,16 @@ func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInf
}
func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
if strings.HasPrefix(info.UpstreamModelName, "agent") {
a.BotType = BotTypeAgent
} else if strings.HasPrefix(info.UpstreamModelName, "workflow") {
a.BotType = BotTypeWorkFlow
} else if strings.HasPrefix(info.UpstreamModelName, "chat") {
a.BotType = BotTypeCompletion
} else {
a.BotType = BotTypeChatFlow
}
//if strings.HasPrefix(info.UpstreamModelName, "agent") {
// a.BotType = BotTypeAgent
//} else if strings.HasPrefix(info.UpstreamModelName, "workflow") {
// a.BotType = BotTypeWorkFlow
//} else if strings.HasPrefix(info.UpstreamModelName, "chat") {
// a.BotType = BotTypeCompletion
//} else {
//}
a.BotType = BotTypeChatFlow
}
func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
......@@ -64,7 +70,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......
relay/channel/gemini/adaptor.go
......@@ -21,6 +21,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -89,7 +95,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......
relay/channel/jina/adaptor.go
......@@ -8,13 +8,21 @@ import (
"net/http"
"one-api/dto"
"one-api/relay/channel"
"one-api/relay/channel/openai"
relaycommon "one-api/relay/common"
"one-api/relay/common_handler"
"one-api/relay/constant"
)
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -43,7 +51,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
return request, nil
}
......@@ -61,9 +69,9 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
if info.RelayMode == constant.RelayModeRerank {
err, usage = JinaRerankHandler(c, resp)
err, usage = common_handler.RerankHandler(c, resp)
} else if info.RelayMode == constant.RelayModeEmbeddings {
err, usage = jinaEmbeddingHandler(c, resp)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/channel/jina/relay-jina.go
package jina
import (
"encoding/json"
"github.com/gin-gonic/gin"
"io"
"net/http"
"one-api/dto"
"one-api/service"
)
func JinaRerankHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
responseBody, err := io.ReadAll(resp.Body)
if err != nil {
return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
}
err = resp.Body.Close()
if err != nil {
return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
}
var jinaResp dto.RerankResponse
err = json.Unmarshal(responseBody, &jinaResp)
if err != nil {
return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
}
jsonResponse, err := json.Marshal(jinaResp)
if err != nil {
return service.OpenAIErrorWrapper(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, &jinaResp.Usage
}
func jinaEmbeddingHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
responseBody, err := io.ReadAll(resp.Body)
if err != nil {
return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
}
err = resp.Body.Close()
if err != nil {
return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
}
var jinaResp dto.OpenAIEmbeddingResponse
err = json.Unmarshal(responseBody, &jinaResp)
if err != nil {
return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
}
jsonResponse, err := json.Marshal(jinaResp)
if err != nil {
return service.OpenAIErrorWrapper(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, &jinaResp.Usage
}
relay/channel/mistral/adaptor.go
......@@ -14,6 +14,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -37,7 +43,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -61,7 +67,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/channel/mokaai/adaptor.go
......@@ -16,6 +16,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -51,7 +57,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -73,13 +79,13 @@ func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, request
}
func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
switch info.RelayMode {
case constant.RelayModeEmbeddings:
err, usage = mokaEmbeddingHandler(c, resp)
default:
// err, usage = mokaHandler(c, resp)
}
return
}
......
relay/channel/ollama/adaptor.go
......@@ -15,6 +15,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -43,7 +49,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -69,7 +75,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.RelayMode == relayconstant.RelayModeEmbeddings {
err, usage = ollamaEmbeddingHandler(c, resp, info.PromptTokens, info.UpstreamModelName, info.RelayMode)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
}
return
......
relay/channel/openai/adaptor.go
......@@ -5,7 +5,6 @@ import (
"encoding/json"
"errors"
"fmt"
"github.com/gin-gonic/gin"
"io"
"mime/multipart"
"net/http"
......@@ -14,13 +13,18 @@ import (
"one-api/dto"
"one-api/relay/channel"
"one-api/relay/channel/ai360"
"one-api/relay/channel/jina"
"one-api/relay/channel/lingyiwanwu"
"one-api/relay/channel/minimax"
"one-api/relay/channel/moonshot"
"one-api/relay/channel/openrouter"
"one-api/relay/channel/xinference"
relaycommon "one-api/relay/common"
"one-api/relay/common_handler"
"one-api/relay/constant"
"one-api/service"
"strings"
"github.com/gin-gonic/gin"
)
type Adaptor struct {
......@@ -28,11 +32,39 @@ type Adaptor struct {
ResponseFormat string
}
func (a *Adaptor) ConvertClaudeRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.ClaudeRequest) (any, error) {
if !strings.Contains(request.Model, "claude") {
return nil, fmt.Errorf("you are using openai channel type with path /v1/messages, only claude model supported convert, but got %s", request.Model)
}
aiRequest, err := service.ClaudeToOpenAIRequest(*request)
if err != nil {
return nil, err
}
if info.SupportStreamOptions {
aiRequest.StreamOptions = &dto.StreamOptions{
IncludeUsage: true,
}
}
return a.ConvertOpenAIRequest(c, info, aiRequest)
}
func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
a.ChannelType = info.ChannelType
// initialize ThinkingContentInfo when thinking_to_content is enabled
if think2Content, ok := info.ChannelSetting[constant2.ChannelSettingThinkingToContent].(bool); ok && think2Content {
info.ThinkingContentInfo = relaycommon.ThinkingContentInfo{
IsFirstThinkingContent: true,
SendLastThinkingContent: false,
HasSentThinkingContent: false,
}
}
}
func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
if info.RelayFormat == relaycommon.RelayFormatClaude {
return fmt.Sprintf("%s/v1/chat/completions", info.BaseUrl), nil
}
if info.RelayMode == constant.RelayModeRealtime {
if strings.HasPrefix(info.BaseUrl, "https://") {
baseUrl := strings.TrimPrefix(info.BaseUrl, "https://")
......@@ -101,14 +133,14 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, header *http.Header, info *
} else {
header.Set("Authorization", "Bearer "+info.ApiKey)
}
//if info.ChannelType == common.ChannelTypeOpenRouter {
// req.Header.Set("HTTP-Referer", "https://github.com/songquanpeng/one-api")
// req.Header.Set("X-Title", "One API")
//}
if info.ChannelType == common.ChannelTypeOpenRouter {
header.Set("HTTP-Referer", "https://github.com/Calcium-Ion/new-api")
header.Set("X-Title", "New API")
}
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -230,12 +262,12 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
case constant.RelayModeImagesGenerations:
err, usage = OpenaiTTSHandler(c, resp, info)
case constant.RelayModeRerank:
err, usage = jina.JinaRerankHandler(c, resp)
err, usage = common_handler.RerankHandler(c, resp)
default:
if info.IsStream {
err, usage = OaiStreamHandler(c, resp, info)
} else {
err, usage = OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = OpenaiHandler(c, resp, info)
}
}
return
......@@ -251,6 +283,10 @@ func (a *Adaptor) GetModelList() []string {
return lingyiwanwu.ModelList
case common.ChannelTypeMiniMax:
return minimax.ModelList
case common.ChannelTypeXinference:
return xinference.ModelList
case common.ChannelTypeOpenRouter:
return openrouter.ModelList
default:
return ModelList
}
......@@ -266,6 +302,10 @@ func (a *Adaptor) GetChannelName() string {
return lingyiwanwu.ChannelName
case common.ChannelTypeMiniMax:
return minimax.ChannelName
case common.ChannelTypeXinference:
return xinference.ChannelName
case common.ChannelTypeOpenRouter:
return openrouter.ChannelName
default:
return ChannelName
}
......
relay/channel/openai/helper.go 0 → 100644
package openai
import (
"encoding/json"
"one-api/common"
"one-api/dto"
relaycommon "one-api/relay/common"
relayconstant "one-api/relay/constant"
"one-api/relay/helper"
"one-api/service"
"strings"
"github.com/gin-gonic/gin"
)
// 辅助函数
func handleStreamFormat(c *gin.Context, info *relaycommon.RelayInfo, data string, forceFormat bool, thinkToContent bool) error {
info.SendResponseCount++
switch info.RelayFormat {
case relaycommon.RelayFormatOpenAI:
return sendStreamData(c, info, data, forceFormat, thinkToContent)
case relaycommon.RelayFormatClaude:
return handleClaudeFormat(c, data, info)
}
return nil
}
func handleClaudeFormat(c *gin.Context, data string, info *relaycommon.RelayInfo) error {
var streamResponse dto.ChatCompletionsStreamResponse
if err := json.Unmarshal(common.StringToByteSlice(data), &streamResponse); err != nil {
return err
}
claudeResponses := service.StreamResponseOpenAI2Claude(&streamResponse, info)
for _, resp := range claudeResponses {
helper.ClaudeData(c, *resp)
}
return nil
}
func processStreamResponse(item string, responseTextBuilder *strings.Builder, toolCount *int) error {
var streamResponse dto.ChatCompletionsStreamResponse
if err := json.Unmarshal(common.StringToByteSlice(item), &streamResponse); err != nil {
return err
}
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Delta.GetContentString())
responseTextBuilder.WriteString(choice.Delta.GetReasoningContent())
if choice.Delta.ToolCalls != nil {
if len(choice.Delta.ToolCalls) > *toolCount {
*toolCount = len(choice.Delta.ToolCalls)
}
for _, tool := range choice.Delta.ToolCalls {
responseTextBuilder.WriteString(tool.Function.Name)
responseTextBuilder.WriteString(tool.Function.Arguments)
}
}
}
return nil
}
func processTokens(relayMode int, streamItems []string, responseTextBuilder *strings.Builder, toolCount *int) error {
streamResp := "[" + strings.Join(streamItems, ",") + "]"
switch relayMode {
case relayconstant.RelayModeChatCompletions:
return processChatCompletions(streamResp, streamItems, responseTextBuilder, toolCount)
case relayconstant.RelayModeCompletions:
return processCompletions(streamResp, streamItems, responseTextBuilder)
}
return nil
}
func processChatCompletions(streamResp string, streamItems []string, responseTextBuilder *strings.Builder, toolCount *int) error {
var streamResponses []dto.ChatCompletionsStreamResponse
if err := json.Unmarshal(common.StringToByteSlice(streamResp), &streamResponses); err != nil {
// 一次性解析失败，逐个解析
common.SysError("error unmarshalling stream response: " + err.Error())
for _, item := range streamItems {
if err := processStreamResponse(item, responseTextBuilder, toolCount); err != nil {
common.SysError("error processing stream response: " + err.Error())
}
}
return nil
}
// 批量处理所有响应
for _, streamResponse := range streamResponses {
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Delta.GetContentString())
responseTextBuilder.WriteString(choice.Delta.GetReasoningContent())
if choice.Delta.ToolCalls != nil {
if len(choice.Delta.ToolCalls) > *toolCount {
*toolCount = len(choice.Delta.ToolCalls)
}
for _, tool := range choice.Delta.ToolCalls {
responseTextBuilder.WriteString(tool.Function.Name)
responseTextBuilder.WriteString(tool.Function.Arguments)
}
}
}
}
return nil
}
func processCompletions(streamResp string, streamItems []string, responseTextBuilder *strings.Builder) error {
var streamResponses []dto.CompletionsStreamResponse
if err := json.Unmarshal(common.StringToByteSlice(streamResp), &streamResponses); err != nil {
// 一次性解析失败，逐个解析
common.SysError("error unmarshalling stream response: " + err.Error())
for _, item := range streamItems {
var streamResponse dto.CompletionsStreamResponse
if err := json.Unmarshal(common.StringToByteSlice(item), &streamResponse); err != nil {
continue
}
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Text)
}
}
return nil
}
// 批量处理所有响应
for _, streamResponse := range streamResponses {
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Text)
}
}
return nil
}
func handleLastResponse(lastStreamData string, responseId *string, createAt *int64,
systemFingerprint *string, model *string, usage **dto.Usage,
containStreamUsage *bool, info *relaycommon.RelayInfo,
shouldSendLastResp *bool) error {
var lastStreamResponse dto.ChatCompletionsStreamResponse
if err := json.Unmarshal(common.StringToByteSlice(lastStreamData), &lastStreamResponse); err != nil {
return err
}
*responseId = lastStreamResponse.Id
*createAt = lastStreamResponse.Created
*systemFingerprint = lastStreamResponse.GetSystemFingerprint()
*model = lastStreamResponse.Model
if service.ValidUsage(lastStreamResponse.Usage) {
*containStreamUsage = true
*usage = lastStreamResponse.Usage
if !info.ShouldIncludeUsage {
*shouldSendLastResp = false
}
}
return nil
}
func handleFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, lastStreamData string,
responseId string, createAt int64, model string, systemFingerprint string,
usage *dto.Usage, containStreamUsage bool) {
switch info.RelayFormat {
case relaycommon.RelayFormatOpenAI:
if info.ShouldIncludeUsage && !containStreamUsage {
response := helper.GenerateFinalUsageResponse(responseId, createAt, model, *usage)
response.SetSystemFingerprint(systemFingerprint)
helper.ObjectData(c, response)
}
helper.Done(c)
case relaycommon.RelayFormatClaude:
var streamResponse dto.ChatCompletionsStreamResponse
if err := json.Unmarshal(common.StringToByteSlice(lastStreamData), &streamResponse); err != nil {
common.SysError("error unmarshalling stream response: " + err.Error())
return
}
if !containStreamUsage {
streamResponse.Usage = usage
}
claudeResponses := service.StreamResponseOpenAI2Claude(&streamResponse, info)
for _, resp := range claudeResponses {
helper.ClaudeData(c, *resp)
}
}
}
relay/channel/openai/relay-openai.go
......@@ -12,7 +12,6 @@ import (
"one-api/constant"
"one-api/dto"
relaycommon "one-api/relay/common"
relayconstant "one-api/relay/constant"
"one-api/relay/helper"
"one-api/service"
"os"
......@@ -66,6 +65,7 @@ func sendStreamData(c *gin.Context, info *relaycommon.RelayInfo, data string, fo
response.Choices[i].Delta.Reasoning = nil
}
info.ThinkingContentInfo.IsFirstThinkingContent = false
info.ThinkingContentInfo.HasSentThinkingContent = true
return helper.ObjectData(c, response)
}
}
......@@ -77,7 +77,8 @@ func sendStreamData(c *gin.Context, info *relaycommon.RelayInfo, data string, fo
// Process each choice
for i, choice := range lastStreamResponse.Choices {
// Handle transition from thinking to content
if hasContent && !info.ThinkingContentInfo.SendLastThinkingContent {
// only send `</think>` tag when previous thinking content has been sent
if hasContent && !info.ThinkingContentInfo.SendLastThinkingContent && info.ThinkingContentInfo.HasSentThinkingContent {
response := lastStreamResponse.Copy()
for j := range response.Choices {
response.Choices[j].Delta.SetContentString("\n</think>\n")
......@@ -88,7 +89,7 @@ func sendStreamData(c *gin.Context, info *relaycommon.RelayInfo, data string, fo
helper.ObjectData(c, response)
}
// Convert reasoning content to regular content
// Convert reasoning content to regular content if any
if len(choice.Delta.GetReasoningContent()) > 0 {
lastStreamResponse.Choices[i].Delta.SetContentString(choice.Delta.GetReasoningContent())
lastStreamResponse.Choices[i].Delta.ReasoningContent = nil
......@@ -137,10 +138,11 @@ func OaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
helper.StreamScannerHandler(c, resp, info, func(data string) bool {
if lastStreamData != "" {
err := sendStreamData(c, info, lastStreamData, forceFormat, thinkToContent)
err := handleStreamFormat(c, info, lastStreamData, forceFormat, thinkToContent)
if err != nil {
common.LogError(c, "streaming error: "+err.Error())
common.SysError("error handling stream format: " + err.Error())
}
info.SetFirstResponseTime()
}
lastStreamData = data
streamItems = append(streamItems, data)
......@@ -172,83 +174,9 @@ func OaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
sendStreamData(c, info, lastStreamData, forceFormat, thinkToContent)
}
// 计算token
streamResp := "[" + strings.Join(streamItems, ",") + "]"
switch info.RelayMode {
case relayconstant.RelayModeChatCompletions:
var streamResponses []dto.ChatCompletionsStreamResponse
err := json.Unmarshal(common.StringToByteSlice(streamResp), &streamResponses)
if err != nil {
// 一次性解析失败，逐个解析
common.SysError("error unmarshalling stream response: " + err.Error())
for _, item := range streamItems {
var streamResponse dto.ChatCompletionsStreamResponse
err := json.Unmarshal(common.StringToByteSlice(item), &streamResponse)
if err == nil {
//if service.ValidUsage(streamResponse.Usage) {
// usage = streamResponse.Usage
//}
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Delta.GetContentString())
// handle both reasoning_content and reasoning
responseTextBuilder.WriteString(choice.Delta.GetReasoningContent())
if choice.Delta.ToolCalls != nil {
if len(choice.Delta.ToolCalls) > toolCount {
toolCount = len(choice.Delta.ToolCalls)
}
for _, tool := range choice.Delta.ToolCalls {
responseTextBuilder.WriteString(tool.Function.Name)
responseTextBuilder.WriteString(tool.Function.Arguments)
}
}
}
}
}
} else {
for _, streamResponse := range streamResponses {
//if service.ValidUsage(streamResponse.Usage) {
// usage = streamResponse.Usage
// containStreamUsage = true
//}
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Delta.GetContentString())
responseTextBuilder.WriteString(choice.Delta.GetReasoningContent()) // This will handle both reasoning_content and reasoning
if choice.Delta.ToolCalls != nil {
if len(choice.Delta.ToolCalls) > toolCount {
toolCount = len(choice.Delta.ToolCalls)
}
for _, tool := range choice.Delta.ToolCalls {
responseTextBuilder.WriteString(tool.Function.Name)
responseTextBuilder.WriteString(tool.Function.Arguments)
}
}
}
}
}
case relayconstant.RelayModeCompletions:
var streamResponses []dto.CompletionsStreamResponse
err := json.Unmarshal(common.StringToByteSlice(streamResp), &streamResponses)
if err != nil {
// 一次性解析失败，逐个解析
common.SysError("error unmarshalling stream response: " + err.Error())
for _, item := range streamItems {
var streamResponse dto.CompletionsStreamResponse
err := json.Unmarshal(common.StringToByteSlice(item), &streamResponse)
if err == nil {
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Text)
}
}
}
} else {
for _, streamResponse := range streamResponses {
for _, choice := range streamResponse.Choices {
responseTextBuilder.WriteString(choice.Text)
}
}
}
// 处理token计算
if err := processTokens(info.RelayMode, streamItems, &responseTextBuilder, &toolCount); err != nil {
common.SysError("error processing tokens: " + err.Error())
}
if !containStreamUsage {
......@@ -262,19 +190,12 @@ func OaiStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.Rel
}
}
if info.ShouldIncludeUsage && !containStreamUsage {
response := helper.GenerateFinalUsageResponse(responseId, createAt, model, *usage)
response.SetSystemFingerprint(systemFingerprint)
helper.ObjectData(c, response)
}
helper.Done(c)
handleFinalResponse(c, info, lastStreamData, responseId, createAt, model, systemFingerprint, usage, containStreamUsage)
//resp.Body.Close()
return nil, usage
}
func OpenaiHandler(c *gin.Context, resp *http.Response, promptTokens int, model string) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
func OpenaiHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
var simpleResponse dto.SimpleResponse
responseBody, err := io.ReadAll(resp.Body)
if err != nil {
......@@ -312,13 +233,13 @@ func OpenaiHandler(c *gin.Context, resp *http.Response, promptTokens int, model
if simpleResponse.Usage.TotalTokens == 0 || (simpleResponse.Usage.PromptTokens == 0 && simpleResponse.Usage.CompletionTokens == 0) {
completionTokens := 0
for _, choice := range simpleResponse.Choices {
ctkm, _ := service.CountTextToken(choice.Message.StringContent()+choice.Message.ReasoningContent+choice.Message.Reasoning, model)
ctkm, _ := service.CountTextToken(choice.Message.StringContent()+choice.Message.ReasoningContent+choice.Message.Reasoning, info.UpstreamModelName)
completionTokens += ctkm
}
simpleResponse.Usage = dto.Usage{
PromptTokens: promptTokens,
PromptTokens: info.PromptTokens,
CompletionTokens: completionTokens,
TotalTokens: promptTokens + completionTokens,
TotalTokens: info.PromptTokens + completionTokens,
}
}
return nil, &simpleResponse.Usage
......
relay/channel/openrouter/adaptor.go deleted 100644 → 0
package openrouter
import (
"errors"
"fmt"
"github.com/gin-gonic/gin"
"io"
"net/http"
"one-api/dto"
"one-api/relay/channel"
"one-api/relay/channel/openai"
relaycommon "one-api/relay/common"
)
type Adaptor struct {
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
}
func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.ImageRequest) (any, error) {
//TODO implement me
return nil, errors.New("not implemented")
}
func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
}
func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
return fmt.Sprintf("%s/v1/chat/completions", info.BaseUrl), nil
}
func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *relaycommon.RelayInfo) error {
channel.SetupApiRequestHeader(info, c, req)
req.Set("Authorization", fmt.Sprintf("Bearer %s", info.ApiKey))
req.Set("HTTP-Referer", "https://github.com/Calcium-Ion/new-api")
req.Set("X-Title", "New API")
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
return request, nil
}
func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
return channel.DoApiRequest(a, c, info, requestBody)
}
func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
return nil, errors.New("not implemented")
}
func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.EmbeddingRequest) (any, error) {
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
}
return
}
func (a *Adaptor) GetModelList() []string {
return ModelList
}
func (a *Adaptor) GetChannelName() string {
return ChannelName
}
relay/channel/palm/adaptor.go
......@@ -15,6 +15,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -38,7 +44,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -54,7 +60,6 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
return channel.DoApiRequest(a, c, info, requestBody)
}
......
relay/channel/perplexity/adaptor.go
......@@ -15,6 +15,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -38,7 +44,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -57,7 +63,6 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
return channel.DoApiRequest(a, c, info, requestBody)
}
......@@ -66,7 +71,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/channel/siliconflow/adaptor.go
......@@ -16,6 +16,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -48,7 +54,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
return request, nil
}
......@@ -72,16 +78,16 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
case constant.RelayModeCompletions:
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
case constant.RelayModeEmbeddings:
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/channel/tencent/adaptor.go
......@@ -23,6 +23,12 @@ type Adaptor struct {
Timestamp int64
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -52,7 +58,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -78,7 +84,6 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
return channel.DoApiRequest(a, c, info, requestBody)
}
......
relay/channel/vertex/adaptor.go
......@@ -38,6 +38,9 @@ type Adaptor struct {
AccountCredentials Credentials
}
func (a *Adaptor) ConvertClaudeRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.ClaudeRequest) (any, error) {
return request, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -119,7 +122,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -175,7 +178,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
case RequestModeGemini:
err, usage = gemini.GeminiChatHandler(c, resp, info)
case RequestModeLlama:
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.OriginModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
}
return
......
relay/channel/vertex/dto.go
package vertex
import (
"one-api/relay/channel/claude"
"one-api/dto"
)
type VertexAIClaudeRequest struct {
AnthropicVersion string `json:"anthropic_version"`
Messages []claude.ClaudeMessage `json:"messages"`
System any `json:"system,omitempty"`
MaxTokens uint `json:"max_tokens,omitempty"`
StopSequences []string `json:"stop_sequences,omitempty"`
Stream bool `json:"stream,omitempty"`
Temperature *float64 `json:"temperature,omitempty"`
TopP float64 `json:"top_p,omitempty"`
TopK int `json:"top_k,omitempty"`
Tools any `json:"tools,omitempty"`
ToolChoice any `json:"tool_choice,omitempty"`
Thinking *claude.Thinking `json:"thinking,omitempty"`
AnthropicVersion string `json:"anthropic_version"`
Messages []dto.ClaudeMessage `json:"messages"`
System any `json:"system,omitempty"`
MaxTokens uint `json:"max_tokens,omitempty"`
StopSequences []string `json:"stop_sequences,omitempty"`
Stream bool `json:"stream,omitempty"`
Temperature *float64 `json:"temperature,omitempty"`
TopP float64 `json:"top_p,omitempty"`
TopK int `json:"top_k,omitempty"`
Tools any `json:"tools,omitempty"`
ToolChoice any `json:"tool_choice,omitempty"`
Thinking *dto.Thinking `json:"thinking,omitempty"`
}
func copyRequest(req *claude.ClaudeRequest, version string) *VertexAIClaudeRequest {
func copyRequest(req *dto.ClaudeRequest, version string) *VertexAIClaudeRequest {
return &VertexAIClaudeRequest{
AnthropicVersion: version,
System: req.System,
......
relay/channel/volcengine/adaptor.go
......@@ -17,6 +17,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -50,7 +56,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -75,10 +81,10 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
case constant.RelayModeEmbeddings:
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/channel/xinference/constant.go 0 → 100644
package xinference
var ModelList = []string{
"bge-reranker-v2-m3",
}
var ChannelName = "xinference"
relay/channel/xunfei/adaptor.go
......@@ -16,6 +16,12 @@ type Adaptor struct {
request *dto.GeneralOpenAIRequest
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -38,7 +44,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -55,7 +61,6 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
// xunfei's request is not http request, so we don't need to do anything here
dummyResp := &http.Response{}
......
relay/channel/zhipu/adaptor.go
......@@ -14,6 +14,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -42,7 +48,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -61,7 +67,6 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
return channel.DoApiRequest(a, c, info, requestBody)
}
......
relay/channel/zhipu_4v/adaptor.go
......@@ -15,6 +15,12 @@ import (
type Adaptor struct {
}
func (a *Adaptor) ConvertClaudeRequest(*gin.Context, *relaycommon.RelayInfo, *dto.ClaudeRequest) (any, error) {
//TODO implement me
panic("implement me")
return nil, nil
}
func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
//TODO implement me
return nil, errors.New("not implemented")
......@@ -39,7 +45,7 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *rel
return nil
}
func (a *Adaptor) ConvertRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
if request == nil {
return nil, errors.New("request is nil")
}
......@@ -58,7 +64,6 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
return nil, errors.New("not implemented")
}
func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
return channel.DoApiRequest(a, c, info, requestBody)
}
......@@ -67,7 +72,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
if info.IsStream {
err, usage = openai.OaiStreamHandler(c, resp, info)
} else {
err, usage = openai.OpenaiHandler(c, resp, info.PromptTokens, info.UpstreamModelName)
err, usage = openai.OpenaiHandler(c, resp, info)
}
return
}
......
relay/claude_handler.go 0 → 100644
package relay
import (
"bytes"
"encoding/json"
"errors"
"fmt"
"github.com/gin-gonic/gin"
"io"
"net/http"
"one-api/common"
"one-api/dto"
relaycommon "one-api/relay/common"
"one-api/relay/helper"
"one-api/service"
"one-api/setting/model_setting"
"strings"
)
func getAndValidateClaudeRequest(c *gin.Context) (textRequest *dto.ClaudeRequest, err error) {
textRequest = &dto.ClaudeRequest{}
err = c.ShouldBindJSON(textRequest)
if err != nil {
return nil, err
}
if textRequest.Messages == nil || len(textRequest.Messages) == 0 {
return nil, errors.New("field messages is required")
}
if textRequest.Model == "" {
return nil, errors.New("field model is required")
}
return textRequest, nil
}
func ClaudeHelper(c *gin.Context) (claudeError *dto.ClaudeErrorWithStatusCode) {
relayInfo := relaycommon.GenRelayInfoClaude(c)
// get & validate textRequest 获取并验证文本请求
textRequest, err := getAndValidateClaudeRequest(c)
if err != nil {
return service.ClaudeErrorWrapperLocal(err, "invalid_claude_request", http.StatusBadRequest)
}
if textRequest.Stream {
relayInfo.IsStream = true
}
err = helper.ModelMappedHelper(c, relayInfo)
if err != nil {
return service.ClaudeErrorWrapperLocal(err, "model_mapped_error", http.StatusInternalServerError)
}
textRequest.Model = relayInfo.UpstreamModelName
promptTokens, err := getClaudePromptTokens(textRequest, relayInfo)
// count messages token error 计算promptTokens错误
if err != nil {
return service.ClaudeErrorWrapperLocal(err, "count_token_messages_failed", http.StatusInternalServerError)
}
priceData, err := helper.ModelPriceHelper(c, relayInfo, promptTokens, int(textRequest.MaxTokens))
if err != nil {
return service.ClaudeErrorWrapperLocal(err, "model_price_error", http.StatusInternalServerError)
}
// pre-consume quota 预消耗配额
preConsumedQuota, userQuota, openaiErr := preConsumeQuota(c, priceData.ShouldPreConsumedQuota, relayInfo)
if openaiErr != nil {
return service.OpenAIErrorToClaudeError(openaiErr)
}
defer func() {
if openaiErr != nil {
returnPreConsumedQuota(c, relayInfo, userQuota, preConsumedQuota)
}
}()
adaptor := GetAdaptor(relayInfo.ApiType)
if adaptor == nil {
return service.ClaudeErrorWrapperLocal(fmt.Errorf("invalid api type: %d", relayInfo.ApiType), "invalid_api_type", http.StatusBadRequest)
}
adaptor.Init(relayInfo)
var requestBody io.Reader
if textRequest.MaxTokens == 0 {
textRequest.MaxTokens = uint(model_setting.GetClaudeSettings().GetDefaultMaxTokens(textRequest.Model))
}
if model_setting.GetClaudeSettings().ThinkingAdapterEnabled &&
strings.HasSuffix(textRequest.Model, "-thinking") {
if textRequest.Thinking == nil {
// 因为BudgetTokens 必须大于1024
if textRequest.MaxTokens < 1280 {
textRequest.MaxTokens = 1280
}
// BudgetTokens 为 max_tokens 的 80%
textRequest.Thinking = &dto.Thinking{
Type: "enabled",
BudgetTokens: int(float64(textRequest.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage),
}
// TODO: 临时处理
// https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#important-considerations-when-using-extended-thinking
textRequest.TopP = 0
textRequest.Temperature = common.GetPointer[float64](1.0)
}
textRequest.Model = strings.TrimSuffix(textRequest.Model, "-thinking")
relayInfo.UpstreamModelName = textRequest.Model
}
convertedRequest, err := adaptor.ConvertClaudeRequest(c, relayInfo, textRequest)
if err != nil {
return service.ClaudeErrorWrapperLocal(err, "convert_request_failed", http.StatusInternalServerError)
}
jsonData, err := json.Marshal(convertedRequest)
if common.DebugEnabled {
println("requestBody: ", string(jsonData))
}
if err != nil {
return service.ClaudeErrorWrapperLocal(err, "json_marshal_failed", http.StatusInternalServerError)
}
requestBody = bytes.NewBuffer(jsonData)
statusCodeMappingStr := c.GetString("status_code_mapping")
var httpResp *http.Response
resp, err := adaptor.DoRequest(c, relayInfo, requestBody)
if err != nil {
return service.ClaudeErrorWrapperLocal(err, "do_request_failed", http.StatusInternalServerError)
}
if resp != nil {
httpResp = resp.(*http.Response)
relayInfo.IsStream = relayInfo.IsStream || strings.HasPrefix(httpResp.Header.Get("Content-Type"), "text/event-stream")
if httpResp.StatusCode != http.StatusOK {
openaiErr = service.RelayErrorHandler(httpResp, false)
// reset status code 重置状态码
service.ResetStatusCode(openaiErr, statusCodeMappingStr)
return service.OpenAIErrorToClaudeError(openaiErr)
}
}
usage, openaiErr := adaptor.DoResponse(c, httpResp, relayInfo)
//log.Printf("usage: %v", usage)
if openaiErr != nil {
// reset status code 重置状态码
service.ResetStatusCode(openaiErr, statusCodeMappingStr)
return service.OpenAIErrorToClaudeError(openaiErr)
}
service.PostClaudeConsumeQuota(c, relayInfo, usage.(*dto.Usage), preConsumedQuota, userQuota, priceData, "")
return nil
}
func getClaudePromptTokens(textRequest *dto.ClaudeRequest, info *relaycommon.RelayInfo) (int, error) {
var promptTokens int
var err error
switch info.RelayMode {
default:
promptTokens, err = service.CountTokenClaudeRequest(*textRequest, info.UpstreamModelName)
}
info.PromptTokens = promptTokens
return promptTokens, err
}
relay/common/relay_info.go
......@@ -15,8 +15,24 @@ import (
type ThinkingContentInfo struct {
IsFirstThinkingContent bool
SendLastThinkingContent bool
HasSentThinkingContent bool
}
const (
LastMessageTypeText = "text"
LastMessageTypeTools = "tools"
)
type ClaudeConvertInfo struct {
LastMessagesType string
Index int
}
const (
RelayFormatOpenAI = "openai"
RelayFormatClaude = "claude"
)
type RelayInfo struct {
ChannelType int
ChannelId int
......@@ -58,7 +74,10 @@ type RelayInfo struct {
UserSetting map[string]interface{}
UserEmail string
UserQuota int
RelayFormat string
SendResponseCount int
ThinkingContentInfo
ClaudeConvertInfo
}
// 定义支持流式选项的通道类型
......@@ -82,6 +101,16 @@ func GenRelayInfoWs(c *gin.Context, ws *websocket.Conn) *RelayInfo {
return info
}
func GenRelayInfoClaude(c *gin.Context) *RelayInfo {
info := GenRelayInfo(c)
info.RelayFormat = RelayFormatClaude
info.ShouldIncludeUsage = false
info.ClaudeConvertInfo = ClaudeConvertInfo{
LastMessagesType: LastMessageTypeText,
}
return info
}
func GenRelayInfo(c *gin.Context) *RelayInfo {
channelType := c.GetInt("channel_type")
channelId := c.GetInt("channel_id")
......@@ -123,6 +152,7 @@ func GenRelayInfo(c *gin.Context) *RelayInfo {
ApiKey: strings.TrimPrefix(c.Request.Header.Get("Authorization"), "Bearer "),
Organization: c.GetString("channel_organization"),
ChannelSetting: channelSetting,
RelayFormat: RelayFormatOpenAI,
ThinkingContentInfo: ThinkingContentInfo{
IsFirstThinkingContent: true,
SendLastThinkingContent: false,
......
relay/common_handler/rerank.go 0 → 100644
package common_handler
import (
"encoding/json"
"github.com/gin-gonic/gin"
"io"
"net/http"
"one-api/dto"
"one-api/service"
)
func RerankHandler(c *gin.Context, resp *http.Response) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
responseBody, err := io.ReadAll(resp.Body)
if err != nil {
return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
}
err = resp.Body.Close()
if err != nil {
return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
}
var jinaResp dto.RerankResponse
err = json.Unmarshal(responseBody, &jinaResp)
if err != nil {
return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
}
jsonResponse, err := json.Marshal(jinaResp)
if err != nil {
return service.OpenAIErrorWrapper(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, &jinaResp.Usage
}
relay/constant/api_type.go
......@@ -31,6 +31,7 @@ const (
APITypeVolcEngine
APITypeBaiduV2
APITypeOpenRouter
APITypeXinference
APITypeDummy // this one is only for count, do not add any channel after this
)
......@@ -89,6 +90,8 @@ func ChannelType2APIType(channelType int) (int, bool) {
apiType = APITypeBaiduV2
case common.ChannelTypeOpenRouter:
apiType = APITypeOpenRouter
case common.ChannelTypeXinference:
apiType = APITypeXinference
}
if apiType == -1 {
return APITypeOpenAI, false
......
relay/helper/common.go
......@@ -19,6 +19,30 @@ func SetEventStreamHeaders(c *gin.Context) {
c.Writer.Header().Set("X-Accel-Buffering", "no")
}
func ClaudeData(c *gin.Context, resp dto.ClaudeResponse) error {
jsonData, err := json.Marshal(resp)
if err != nil {
common.SysError("error marshalling stream response: " + err.Error())
} else {
c.Render(-1, common.CustomEvent{Data: fmt.Sprintf("event: %s\n", resp.Type)})
c.Render(-1, common.CustomEvent{Data: "data: " + string(jsonData)})
}
if flusher, ok := c.Writer.(http.Flusher); ok {
flusher.Flush()
} else {
return errors.New("streaming error: flusher not found")
}
return nil
}
func ClaudeChunkData(c *gin.Context, resp dto.ClaudeResponse, data string) {
c.Render(-1, common.CustomEvent{Data: fmt.Sprintf("event: %s\n", resp.Type)})
c.Render(-1, common.CustomEvent{Data: fmt.Sprintf("data: %s\n", data)})
if flusher, ok := c.Writer.(http.Flusher); ok {
flusher.Flush()
}
}
func StringData(c *gin.Context, str string) error {
//str = strings.TrimPrefix(str, "data: ")
//str = strings.TrimSuffix(str, "\r")
......
relay/helper/price.go
......@@ -16,6 +16,7 @@ type PriceData struct {
CacheRatio float64
GroupRatio float64
UsePrice bool
CacheCreationRatio float64
ShouldPreConsumedQuota int
}
......@@ -26,6 +27,7 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
var modelRatio float64
var completionRatio float64
var cacheRatio float64
var cacheCreationRatio float64
if !usePrice {
preConsumedTokens := common.PreConsumedQuota
if maxTokens != 0 {
......@@ -42,6 +44,7 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
}
completionRatio = operation_setting.GetCompletionRatio(info.OriginModelName)
cacheRatio, _ = operation_setting.GetCacheRatio(info.OriginModelName)
cacheCreationRatio, _ = operation_setting.GetCreateCacheRatio(info.OriginModelName)
ratio := modelRatio * groupRatio
preConsumedQuota = int(float64(preConsumedTokens) * ratio)
} else {
......@@ -54,6 +57,7 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
GroupRatio: groupRatio,
UsePrice: usePrice,
CacheRatio: cacheRatio,
CacheCreationRatio: cacheCreationRatio,
ShouldPreConsumedQuota: preConsumedQuota,
}, nil
}
relay/relay-text.go
......@@ -160,7 +160,7 @@ func TextHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) {
}
requestBody = bytes.NewBuffer(body)
} else {
convertedRequest, err := adaptor.ConvertRequest(c, relayInfo, textRequest)
convertedRequest, err := adaptor.ConvertOpenAIRequest(c, relayInfo, textRequest)
if err != nil {
return service.OpenAIErrorWrapperLocal(err, "convert_request_failed", http.StatusInternalServerError)
}
......@@ -168,6 +168,9 @@ func TextHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) {
if err != nil {
return service.OpenAIErrorWrapperLocal(err, "json_marshal_failed", http.StatusInternalServerError)
}
if common.DebugEnabled {
println("requestBody: ", string(jsonData))
}
requestBody = bytes.NewBuffer(jsonData)
}
......
relay/relay_adaptor.go
......@@ -18,7 +18,6 @@ import (
"one-api/relay/channel/mokaai"
"one-api/relay/channel/ollama"
"one-api/relay/channel/openai"
"one-api/relay/channel/openrouter"
"one-api/relay/channel/palm"
"one-api/relay/channel/perplexity"
"one-api/relay/channel/siliconflow"
......@@ -34,8 +33,6 @@ import (
func GetAdaptor(apiType int) channel.Adaptor {
switch apiType {
//case constant.APITypeAIProxyLibrary:
// return &aiproxy.Adaptor{}
case constant.APITypeAli:
return &ali.Adaptor{}
case constant.APITypeAnthropic:
......@@ -85,7 +82,9 @@ func GetAdaptor(apiType int) channel.Adaptor {
case constant.APITypeBaiduV2:
return &baidu_v2.Adaptor{}
case constant.APITypeOpenRouter:
return &openrouter.Adaptor{}
return &openai.Adaptor{}
case constant.APITypeXinference:
return &openai.Adaptor{}
}
return nil
}
......
router/relay-router.go
......@@ -35,6 +35,7 @@ func SetRelayRouter(router *gin.Engine) {
//http router
httpRouter := relayV1Router.Group("")
httpRouter.Use(middleware.Distribute())
httpRouter.POST("/messages", controller.RelayClaude)
httpRouter.POST("/completions", controller.Relay)
httpRouter.POST("/chat/completions", controller.Relay)
httpRouter.POST("/edits", controller.Relay)
......
service/convert.go 0 → 100644
package service
import (
"encoding/json"
"fmt"
"one-api/common"
"one-api/dto"
relaycommon "one-api/relay/common"
)
func ClaudeToOpenAIRequest(claudeRequest dto.ClaudeRequest) (*dto.GeneralOpenAIRequest, error) {
openAIRequest := dto.GeneralOpenAIRequest{
Model: claudeRequest.Model,
MaxTokens: claudeRequest.MaxTokens,
Temperature: claudeRequest.Temperature,
TopP: claudeRequest.TopP,
Stream: claudeRequest.Stream,
}
// Convert stop sequences
if len(claudeRequest.StopSequences) == 1 {
openAIRequest.Stop = claudeRequest.StopSequences[0]
} else if len(claudeRequest.StopSequences) > 1 {
openAIRequest.Stop = claudeRequest.StopSequences
}
// Convert tools
tools, _ := common.Any2Type[[]dto.Tool](claudeRequest.Tools)
openAITools := make([]dto.ToolCallRequest, 0)
for _, claudeTool := range tools {
openAITool := dto.ToolCallRequest{
Type: "function",
Function: dto.FunctionRequest{
Name: claudeTool.Name,
Description: claudeTool.Description,
Parameters: claudeTool.InputSchema,
},
}
openAITools = append(openAITools, openAITool)
}
openAIRequest.Tools = openAITools
// Convert messages
openAIMessages := make([]dto.Message, 0)
// Add system message if present
if claudeRequest.System != nil {
if claudeRequest.IsStringSystem() {
openAIMessage := dto.Message{
Role: "system",
}
openAIMessage.SetStringContent(claudeRequest.GetStringSystem())
openAIMessages = append(openAIMessages, openAIMessage)
} else {
systems := claudeRequest.ParseSystem()
if len(systems) > 0 {
systemStr := ""
openAIMessage := dto.Message{
Role: "system",
}
for _, system := range systems {
systemStr += system.Type
}
openAIMessage.SetStringContent(systemStr)
openAIMessages = append(openAIMessages, openAIMessage)
}
}
}
for _, claudeMessage := range claudeRequest.Messages {
openAIMessage := dto.Message{
Role: claudeMessage.Role,
}
//log.Printf("claudeMessage.Content: %v", claudeMessage.Content)
if claudeMessage.IsStringContent() {
openAIMessage.SetStringContent(claudeMessage.GetStringContent())
} else {
content, err := claudeMessage.ParseContent()
if err != nil {
return nil, err
}
contents := content
var toolCalls []dto.ToolCallRequest
mediaMessages := make([]dto.MediaContent, 0, len(contents))
for _, mediaMsg := range contents {
switch mediaMsg.Type {
case "text":
message := dto.MediaContent{
Type: "text",
Text: mediaMsg.GetText(),
}
mediaMessages = append(mediaMessages, message)
case "image":
// Handle image conversion (base64 to URL or keep as is)
imageData := fmt.Sprintf("data:%s;base64,%s", mediaMsg.Source.MediaType, mediaMsg.Source.Data)
//textContent += fmt.Sprintf("[Image: %s]", imageData)
mediaMessage := dto.MediaContent{
Type: "image_url",
ImageUrl: &dto.MessageImageUrl{Url: imageData},
}
mediaMessages = append(mediaMessages, mediaMessage)
case "tool_use":
toolCall := dto.ToolCallRequest{
ID: mediaMsg.Id,
Type: "function",
Function: dto.FunctionRequest{
Name: mediaMsg.Name,
Arguments: toJSONString(mediaMsg.Input),
},
}
toolCalls = append(toolCalls, toolCall)
case "tool_result":
// Add tool result as a separate message
oaiToolMessage := dto.Message{
Role: "tool",
Name: &mediaMsg.Name,
ToolCallId: mediaMsg.ToolUseId,
}
//oaiToolMessage.SetStringContent(*mediaMsg.GetMediaContent().Text)
if mediaMsg.IsStringContent() {
oaiToolMessage.SetStringContent(mediaMsg.GetStringContent())
} else {
mediaContents := mediaMsg.ParseMediaContent()
if len(mediaContents) > 0 && mediaContents[0].Text != nil {
oaiToolMessage.SetStringContent(*mediaContents[0].Text)
}
}
openAIMessages = append(openAIMessages, oaiToolMessage)
}
}
if len(mediaMessages) > 0 {
openAIMessage.SetMediaContent(mediaMessages)
}
if len(toolCalls) > 0 {
openAIMessage.SetToolCalls(toolCalls)
}
}
if len(openAIMessage.ParseContent()) > 0 {
openAIMessages = append(openAIMessages, openAIMessage)
}
}
openAIRequest.Messages = openAIMessages
return &openAIRequest, nil
}
func OpenAIErrorToClaudeError(openAIError *dto.OpenAIErrorWithStatusCode) *dto.ClaudeErrorWithStatusCode {
claudeError := dto.ClaudeError{
Type: "new_api_error",
Message: openAIError.Error.Message,
}
return &dto.ClaudeErrorWithStatusCode{
Error: claudeError,
StatusCode: openAIError.StatusCode,
}
}
func ClaudeErrorToOpenAIError(claudeError *dto.ClaudeErrorWithStatusCode) *dto.OpenAIErrorWithStatusCode {
openAIError := dto.OpenAIError{
Message: claudeError.Error.Message,
Type: "new_api_error",
}
return &dto.OpenAIErrorWithStatusCode{
Error: openAIError,
StatusCode: claudeError.StatusCode,
}
}
func generateStopBlock(index int) *dto.ClaudeResponse {
return &dto.ClaudeResponse{
Type: "content_block_stop",
Index: common.GetPointer[int](index),
}
}
func StreamResponseOpenAI2Claude(openAIResponse *dto.ChatCompletionsStreamResponse, info *relaycommon.RelayInfo) []*dto.ClaudeResponse {
var claudeResponses []*dto.ClaudeResponse
if info.SendResponseCount == 1 {
msg := &dto.ClaudeMediaMessage{
Id: openAIResponse.Id,
Model: openAIResponse.Model,
Type: "message",
Role: "assistant",
Usage: &dto.ClaudeUsage{
InputTokens: info.PromptTokens,
OutputTokens: 0,
},
}
msg.SetContent(make([]any, 0))
claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
Type: "message_start",
Message: msg,
})
claudeResponses = append(claudeResponses)
//claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
// Type: "ping",
//})
if openAIResponse.IsToolCall() {
resp := &dto.ClaudeResponse{
Type: "content_block_start",
ContentBlock: &dto.ClaudeMediaMessage{
Id: openAIResponse.GetFirstToolCall().ID,
Type: "tool_use",
Name: openAIResponse.GetFirstToolCall().Function.Name,
},
}
resp.SetIndex(0)
claudeResponses = append(claudeResponses, resp)
} else {
resp := &dto.ClaudeResponse{
Type: "content_block_start",
ContentBlock: &dto.ClaudeMediaMessage{
Type: "text",
Text: common.GetPointer[string](""),
},
}
resp.SetIndex(0)
claudeResponses = append(claudeResponses, resp)
}
return claudeResponses
}
if len(openAIResponse.Choices) == 0 {
// no choices
// TODO: handle this case
return claudeResponses
} else {
chosenChoice := openAIResponse.Choices[0]
if chosenChoice.FinishReason != nil && *chosenChoice.FinishReason != "" {
// should be done
claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
if openAIResponse.Usage != nil {
claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
Type: "message_delta",
Usage: &dto.ClaudeUsage{
InputTokens: openAIResponse.Usage.PromptTokens,
OutputTokens: openAIResponse.Usage.CompletionTokens,
},
Delta: &dto.ClaudeMediaMessage{
StopReason: common.GetPointer[string](stopReasonOpenAI2Claude(*chosenChoice.FinishReason)),
},
})
}
claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
Type: "message_stop",
})
} else {
var claudeResponse dto.ClaudeResponse
claudeResponse.SetIndex(0)
claudeResponse.Type = "content_block_delta"
if len(chosenChoice.Delta.ToolCalls) > 0 {
if info.ClaudeConvertInfo.LastMessagesType == relaycommon.LastMessageTypeText {
claudeResponses = append(claudeResponses, generateStopBlock(info.ClaudeConvertInfo.Index))
info.ClaudeConvertInfo.Index++
claudeResponses = append(claudeResponses, &dto.ClaudeResponse{
Index: &info.ClaudeConvertInfo.Index,
Type: "content_block_start",
ContentBlock: &dto.ClaudeMediaMessage{
Id: openAIResponse.GetFirstToolCall().ID,
Type: "tool_use",
Name: openAIResponse.GetFirstToolCall().Function.Name,
Input: map[string]interface{}{},
},
})
}
info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeTools
// tools delta
claudeResponse.Delta = &dto.ClaudeMediaMessage{
Type: "input_json_delta",
PartialJson: &chosenChoice.Delta.ToolCalls[0].Function.Arguments,
}
} else {
info.ClaudeConvertInfo.LastMessagesType = relaycommon.LastMessageTypeText
// text delta
claudeResponse.Delta = &dto.ClaudeMediaMessage{
Type: "text_delta",
Text: common.GetPointer[string](chosenChoice.Delta.GetContentString()),
}
}
claudeResponse.Index = &info.ClaudeConvertInfo.Index
claudeResponses = append(claudeResponses, &claudeResponse)
}
}
return claudeResponses
}
func ResponseOpenAI2Claude(openAIResponse *dto.OpenAITextResponse, info *relaycommon.RelayInfo) *dto.ClaudeResponse {
var stopReason string
contents := make([]dto.ClaudeMediaMessage, 0)
claudeResponse := &dto.ClaudeResponse{
Id: openAIResponse.Id,
Type: "message",
Role: "assistant",
Model: openAIResponse.Model,
}
for _, choice := range openAIResponse.Choices {
stopReason = stopReasonOpenAI2Claude(choice.FinishReason)
claudeContent := dto.ClaudeMediaMessage{}
if choice.FinishReason == "tool_calls" {
claudeContent.Type = "tool_use"
claudeContent.Id = choice.Message.ToolCallId
claudeContent.Name = choice.Message.ParseToolCalls()[0].Function.Name
var mapParams map[string]interface{}
if err := json.Unmarshal([]byte(choice.Message.ParseToolCalls()[0].Function.Arguments), &mapParams); err == nil {
claudeContent.Input = mapParams
} else {
claudeContent.Input = choice.Message.ParseToolCalls()[0].Function.Arguments
}
} else {
claudeContent.Type = "text"
claudeContent.SetText(choice.Message.StringContent())
}
contents = append(contents, claudeContent)
}
claudeResponse.Content = contents
claudeResponse.StopReason = stopReason
claudeResponse.Usage = &dto.ClaudeUsage{
InputTokens: openAIResponse.PromptTokens,
OutputTokens: openAIResponse.CompletionTokens,
}
return claudeResponse
}
func stopReasonOpenAI2Claude(reason string) string {
switch reason {
case "stop":
return "end_turn"
case "stop_sequence":
return "stop_sequence"
case "max_tokens":
return "max_tokens"
case "tool_calls":
return "tool_use"
default:
return reason
}
}
func toJSONString(v interface{}) string {
b, err := json.Marshal(v)
if err != nil {
return "{}"
}
return string(b)
}
service/error.go
......@@ -50,6 +50,30 @@ func OpenAIErrorWrapperLocal(err error, code string, statusCode int) *dto.OpenAI
return openaiErr
}
func ClaudeErrorWrapper(err error, code string, statusCode int) *dto.ClaudeErrorWithStatusCode {
text := err.Error()
lowerText := strings.ToLower(text)
if strings.Contains(lowerText, "post") || strings.Contains(lowerText, "dial") || strings.Contains(lowerText, "http") {
common.SysLog(fmt.Sprintf("error: %s", text))
text = "请求上游地址失败"
}
claudeError := dto.ClaudeError{
Message: text,
Type: "new_api_error",
//Code: code,
}
return &dto.ClaudeErrorWithStatusCode{
Error: claudeError,
StatusCode: statusCode,
}
}
func ClaudeErrorWrapperLocal(err error, code string, statusCode int) *dto.ClaudeErrorWithStatusCode {
claudeErr := ClaudeErrorWrapper(err, code, statusCode)
claudeErr.LocalError = true
return claudeErr
}
func RelayErrorHandler(resp *http.Response, showBodyWhenFail bool) (errWithStatusCode *dto.OpenAIErrorWithStatusCode) {
errWithStatusCode = &dto.OpenAIErrorWithStatusCode{
StatusCode: resp.StatusCode,
......
service/log_info_generate.go
......@@ -53,3 +53,12 @@ func GenerateAudioOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
info["audio_completion_ratio"] = audioCompletionRatio
return info
}
func GenerateClaudeOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, modelRatio, groupRatio, completionRatio float64,
cacheTokens int, cacheRatio float64, cacheCreationTokens int, cacheCreationRatio float64, modelPrice float64) map[string]interface{} {
info := GenerateTextOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio, cacheTokens, cacheRatio, modelPrice)
info["claude"] = true
info["cache_creation_tokens"] = cacheCreationTokens
info["cache_creation_ratio"] = cacheCreationRatio
return info
}
service/quota.go
......@@ -194,6 +194,75 @@ func PostWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, mod
tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, relayInfo.Group, other)
}
func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
usage *dto.Usage, preConsumedQuota int, userQuota int, priceData helper.PriceData, extraContent string) {
useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
promptTokens := usage.PromptTokens
completionTokens := usage.CompletionTokens
modelName := relayInfo.OriginModelName
tokenName := ctx.GetString("token_name")
completionRatio := priceData.CompletionRatio
modelRatio := priceData.ModelRatio
groupRatio := priceData.GroupRatio
modelPrice := priceData.ModelPrice
cacheRatio := priceData.CacheRatio
cacheTokens := usage.PromptTokensDetails.CachedTokens
cacheCreationRatio := priceData.CacheCreationRatio
cacheCreationTokens := usage.PromptTokensDetails.CachedCreationTokens
calculateQuota := 0.0
if !priceData.UsePrice {
calculateQuota = float64(promptTokens)
calculateQuota += float64(cacheTokens) * cacheRatio
calculateQuota += float64(cacheCreationTokens) * cacheCreationRatio
calculateQuota += float64(completionTokens) * completionRatio
calculateQuota = calculateQuota * groupRatio * modelRatio
} else {
calculateQuota = modelPrice * common.QuotaPerUnit * groupRatio
}
if modelRatio != 0 && calculateQuota <= 0 {
calculateQuota = 1
}
quota := int(calculateQuota)
totalTokens := promptTokens + completionTokens
var logContent string
// record all the consume log even if quota is 0
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
logContent += fmt.Sprintf("（可能是上游出错）")
common.LogError(ctx, fmt.Sprintf("total tokens is 0, cannot consume quota, userId %d, channelId %d, "+
"tokenId %d, model %s， pre-consumed quota %d", relayInfo.UserId, relayInfo.ChannelId, relayInfo.TokenId, modelName, preConsumedQuota))
} else {
//if sensitiveResp != nil {
// logContent += fmt.Sprintf("，敏感词：%s", strings.Join(sensitiveResp.SensitiveWords, ", "))
//}
quotaDelta := quota - preConsumedQuota
if quotaDelta != 0 {
err := PostConsumeQuota(relayInfo, quotaDelta, preConsumedQuota, true)
if err != nil {
common.LogError(ctx, "error consuming token remain quota: "+err.Error())
}
}
model.UpdateUserUsedQuotaAndRequestCount(relayInfo.UserId, quota)
model.UpdateChannelUsedQuota(relayInfo.ChannelId, quota)
}
other := GenerateClaudeOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio,
cacheTokens, cacheRatio, cacheCreationTokens, cacheCreationRatio, modelPrice)
model.RecordConsumeLog(ctx, relayInfo.UserId, relayInfo.ChannelId, promptTokens, completionTokens, modelName,
tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, relayInfo.Group, other)
}
func PostAudioConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
usage *dto.Usage, preConsumedQuota int, userQuota int, priceData helper.PriceData, extraContent string) {
......
service/token_counter.go
package service
import (
"encoding/json"
"errors"
"fmt"
"image"
......@@ -192,6 +193,110 @@ func CountTokenChatRequest(info *relaycommon.RelayInfo, request dto.GeneralOpenA
return tkm, nil
}
func CountTokenClaudeRequest(request dto.ClaudeRequest, model string) (int, error) {
tkm := 0
// Count tokens in messages
msgTokens, err := CountTokenClaudeMessages(request.Messages, model, request.Stream)
if err != nil {
return 0, err
}
tkm += msgTokens
// Count tokens in system message
if request.System != "" {
systemTokens, err := CountTokenInput(request.System, model)
if err != nil {
return 0, err
}
tkm += systemTokens
}
if request.Tools != nil {
// check is array
if tools, ok := request.Tools.([]any); ok {
if len(tools) > 0 {
parsedTools, err1 := common.Any2Type[[]dto.Tool](request.Tools)
if err1 != nil {
return 0, fmt.Errorf("tools: Input should be a valid list: %v", err)
}
toolTokens, err2 := CountTokenClaudeTools(parsedTools, model)
if err2 != nil {
return 0, fmt.Errorf("tools: %v", err)
}
tkm += toolTokens
}
} else {
return 0, errors.New("tools: Input should be a valid list")
}
}
return tkm, nil
}
func CountTokenClaudeMessages(messages []dto.ClaudeMessage, model string, stream bool) (int, error) {
tokenEncoder := getTokenEncoder(model)
tokenNum := 0
for _, message := range messages {
// Count tokens for role
tokenNum += getTokenNum(tokenEncoder, message.Role)
if message.IsStringContent() {
tokenNum += getTokenNum(tokenEncoder, message.GetStringContent())
} else {
content, err := message.ParseContent()
if err != nil {
return 0, err
}
for _, mediaMessage := range content {
switch mediaMessage.Type {
case "text":
tokenNum += getTokenNum(tokenEncoder, mediaMessage.GetText())
case "image":
//imageTokenNum, err := getClaudeImageToken(mediaMsg.Source, model, stream)
//if err != nil {
// return 0, err
//}
tokenNum += 1000
case "tool_use":
tokenNum += getTokenNum(tokenEncoder, mediaMessage.Name)
inputJSON, _ := json.Marshal(mediaMessage.Input)
tokenNum += getTokenNum(tokenEncoder, string(inputJSON))
case "tool_result":
contentJSON, _ := json.Marshal(mediaMessage.Content)
tokenNum += getTokenNum(tokenEncoder, string(contentJSON))
}
}
}
}
// Add a constant for message formatting (this may need adjustment based on Claude's exact formatting)
tokenNum += len(messages) * 2 // Assuming 2 tokens per message for formatting
return tokenNum, nil
}
func CountTokenClaudeTools(tools []dto.Tool, model string) (int, error) {
tokenEncoder := getTokenEncoder(model)
tokenNum := 0
for _, tool := range tools {
tokenNum += getTokenNum(tokenEncoder, tool.Name)
tokenNum += getTokenNum(tokenEncoder, tool.Description)
schemaJSON, err := json.Marshal(tool.InputSchema)
if err != nil {
return 0, errors.New(fmt.Sprintf("marshal_tool_schema_fail: %s", err.Error()))
}
tokenNum += getTokenNum(tokenEncoder, string(schemaJSON))
}
// Add a constant for tool formatting (this may need adjustment based on Claude's exact formatting)
tokenNum += len(tools) * 3 // Assuming 3 tokens per tool for formatting
return tokenNum, nil
}
func CountTokenRealtime(info *relaycommon.RelayInfo, request dto.RealtimeEvent, model string) (int, int, error) {
audioToken := 0
textToken := 0
......
setting/operation_setting/cache_ratio.go
......@@ -7,26 +7,45 @@ import (
)
var defaultCacheRatio = map[string]float64{
"gpt-4": 0.5,
"o1": 0.5,
"o1-2024-12-17": 0.5,
"o1-preview-2024-09-12": 0.5,
"o1-preview": 0.5,
"o1-mini-2024-09-12": 0.5,
"o1-mini": 0.5,
"gpt-4o-2024-11-20": 0.5,
"gpt-4o-2024-08-06": 0.5,
"gpt-4o": 0.5,
"gpt-4o-mini-2024-07-18": 0.5,
"gpt-4o-mini": 0.5,
"gpt-4o-realtime-preview": 0.5,
"gpt-4o-mini-realtime-preview": 0.5,
"deepseek-chat": 0.25,
"deepseek-reasoner": 0.25,
"deepseek-coder": 0.25,
"gpt-4": 0.5,
"o1": 0.5,
"o1-2024-12-17": 0.5,
"o1-preview-2024-09-12": 0.5,
"o1-preview": 0.5,
"o1-mini-2024-09-12": 0.5,
"o1-mini": 0.5,
"gpt-4o-2024-11-20": 0.5,
"gpt-4o-2024-08-06": 0.5,
"gpt-4o": 0.5,
"gpt-4o-mini-2024-07-18": 0.5,
"gpt-4o-mini": 0.5,
"gpt-4o-realtime-preview": 0.5,
"gpt-4o-mini-realtime-preview": 0.5,
"deepseek-chat": 0.25,
"deepseek-reasoner": 0.25,
"deepseek-coder": 0.25,
"claude-3-sonnet-20240229": 0.1,
"claude-3-opus-20240229": 0.1,
"claude-3-haiku-20240307": 0.1,
"claude-3-5-haiku-20241022": 0.1,
"claude-3-5-sonnet-20240620": 0.1,
"claude-3-5-sonnet-20241022": 0.1,
"claude-3-7-sonnet-20250219": 0.1,
"claude-3-7-sonnet-20250219-thinking": 0.1,
}
var defaultCreateCacheRatio = map[string]float64{}
var defaultCreateCacheRatio = map[string]float64{
"claude-3-sonnet-20240229": 1.25,
"claude-3-opus-20240229": 1.25,
"claude-3-haiku-20240307": 1.25,
"claude-3-5-haiku-20241022": 1.25,
"claude-3-5-sonnet-20240620": 1.25,
"claude-3-5-sonnet-20241022": 1.25,
"claude-3-7-sonnet-20250219": 1.25,
"claude-3-7-sonnet-20250219-thinking": 1.25,
}
//var defaultCreateCacheRatio = map[string]float64{}
var cacheRatioMap map[string]float64
var cacheRatioMapMutex sync.RWMutex
......@@ -69,16 +88,10 @@ func GetCacheRatio(name string) (float64, bool) {
return ratio, true
}
// DefaultCacheRatio2JSONString converts the default cache ratio map to a JSON string
func DefaultCacheRatio2JSONString() string {
jsonBytes, err := json.Marshal(defaultCacheRatio)
if err != nil {
common.SysError("error marshalling default cache ratio: " + err.Error())
func GetCreateCacheRatio(name string) (float64, bool) {
ratio, ok := defaultCreateCacheRatio[name]
if !ok {
return 1.25, false // Default to 1.25 if not found
}
return string(jsonBytes)
}
// GetDefaultCacheRatioMap returns the default cache ratio map
func GetDefaultCacheRatioMap() map[string]float64 {
return defaultCacheRatio
return ratio, true
}
web/src/components/LogsTable.js
......@@ -26,8 +26,14 @@ import {
} from '@douyinfe/semi-ui';
import { ITEMS_PER_PAGE } from '../constants';
import {
renderAudioModelPrice, renderGroup,
renderModelPrice, renderModelPriceSimple,
renderAudioModelPrice,
renderClaudeLogContent,
renderClaudeModelPrice,
renderClaudeModelPriceSimple,
renderGroup,
renderLogContent,
renderModelPrice,
renderModelPriceSimple,
renderNumber,
renderQuota,
stringToColor
......@@ -564,13 +570,23 @@ const LogsTable = () => {
);
}
let content = renderModelPriceSimple(
other.model_ratio,
other.model_price,
other.group_ratio,
other.cache_tokens || 0,
other.cache_ratio || 1.0,
);
let content = other?.claude
? renderClaudeModelPriceSimple(
other.model_ratio,
other.model_price,
other.group_ratio,
other.cache_tokens || 0,
other.cache_ratio || 1.0,
other.cache_creation_tokens || 0,
other.cache_creation_ratio || 1.0,
)
: renderModelPriceSimple(
other.model_ratio,
other.model_price,
other.group_ratio,
other.cache_tokens || 0,
other.cache_ratio || 1.0,
);
return (
<Paragraph
ellipsis={{
......@@ -818,10 +834,34 @@ const LogsTable = () => {
value: other.cache_tokens,
});
}
expandDataLocal.push({
key: t('日志详情'),
value: logs[i].content,
});
if (other?.cache_creation_tokens > 0) {
expandDataLocal.push({
key: t('缓存创建 Tokens'),
value: other.cache_creation_tokens,
});
}
if (logs[i].type === 2) {
expandDataLocal.push({
key: t('日志详情'),
value: other?.claude
? renderClaudeLogContent(
other?.model_ratio,
other.completion_ratio,
other.model_price,
other.group_ratio,
other.user_group_ratio,
other.cache_ratio || 1.0,
other.cache_creation_ratio || 1.0
)
: renderLogContent(
other?.model_ratio,
other.completion_ratio,
other.model_price,
other.group_ratio,
other.user_group_ratio
),
});
}
if (logs[i].type === 2) {
let modelMapped = other?.is_model_mapped && other?.upstream_model_name && other?.upstream_model_name !== '';
if (modelMapped) {
......@@ -850,6 +890,19 @@ const LogsTable = () => {
other?.cache_tokens || 0,
other?.cache_ratio || 1.0,
);
} else if (other?.claude) {
content = renderClaudeModelPrice(
logs[i].prompt_tokens,
logs[i].completion_tokens,
other.model_ratio,
other.model_price,
other.completion_ratio,
other.group_ratio,
other.cache_tokens || 0,
other.cache_ratio || 1.0,
other.cache_creation_tokens || 0,
other.cache_creation_ratio || 1.0,
);
} else {
content = renderModelPrice(
logs[i].prompt_tokens,
......
web/src/components/SafetySetting.js deleted 100644 → 0
import React, { useEffect, useState } from 'react';
import {
Button,
Divider,
Form,
Grid,
Header,
Message,
Modal,
} from 'semantic-ui-react';
import { API, removeTrailingSlash, showError, verifyJSON } from '../helpers';
import { useTheme } from '../context/Theme';
const SafetySetting = () => {
let [inputs, setInputs] = useState({
PasswordLoginEnabled: '',
PasswordRegisterEnabled: '',
EmailVerificationEnabled: '',
GitHubOAuthEnabled: '',
GitHubClientId: '',
GitHubClientSecret: '',
Notice: '',
SMTPServer: '',
SMTPPort: '',
SMTPAccount: '',
SMTPFrom: '',
SMTPToken: '',
ServerAddress: '',
WorkerUrl: '',
WorkerValidKey: '',
EpayId: '',
EpayKey: '',
Price: 7.3,
MinTopUp: 1,
TopupGroupRatio: '',
PayAddress: '',
CustomCallbackAddress: '',
Footer: '',
WeChatAuthEnabled: '',
WeChatServerAddress: '',
WeChatServerToken: '',
WeChatAccountQRCodeImageURL: '',
TurnstileCheckEnabled: '',
TurnstileSiteKey: '',
TurnstileSecretKey: '',
RegisterEnabled: '',
EmailDomainRestrictionEnabled: '',
EmailAliasRestrictionEnabled: '',
SMTPSSLEnabled: '',
EmailDomainWhitelist: [],
// telegram login
TelegramOAuthEnabled: '',
TelegramBotToken: '',
TelegramBotName: '',
});
const [originInputs, setOriginInputs] = useState({});
let [loading, setLoading] = useState(false);
const [EmailDomainWhitelist, setEmailDomainWhitelist] = useState([]);
const [restrictedDomainInput, setRestrictedDomainInput] = useState('');
const [showPasswordWarningModal, setShowPasswordWarningModal] =
useState(false);
const theme = useTheme();
const isDark = theme === 'dark';
const getOptions = async () => {
const res = await API.get('/api/option/');
const { success, message, data } = res.data;
if (success) {
let newInputs = {};
data.forEach((item) => {
if (item.key === 'TopupGroupRatio') {
item.value = JSON.stringify(JSON.parse(item.value), null, 2);
}
newInputs[item.key] = item.value;
});
setInputs({
...newInputs,
EmailDomainWhitelist: newInputs.EmailDomainWhitelist.split(','),
});
setOriginInputs(newInputs);
setEmailDomainWhitelist(
newInputs.EmailDomainWhitelist.split(',').map((item) => {
return { key: item, text: item, value: item };
}),
);
} else {
showError(message);
}
};
useEffect(() => {
getOptions().then();
}, []);
useEffect(() => {}, [inputs.EmailDomainWhitelist]);
const updateOption = async (key, value) => {
setLoading(true);
switch (key) {
case 'PasswordLoginEnabled':
case 'PasswordRegisterEnabled':
case 'EmailVerificationEnabled':
case 'GitHubOAuthEnabled':
case 'WeChatAuthEnabled':
case 'TelegramOAuthEnabled':
case 'TurnstileCheckEnabled':
case 'EmailDomainRestrictionEnabled':
case 'EmailAliasRestrictionEnabled':
case 'SMTPSSLEnabled':
case 'RegisterEnabled':
value = inputs[key] === 'true' ? 'false' : 'true';
break;
default:
break;
}
const res = await API.put('/api/option/', {
key,
value,
});
const { success, message } = res.data;
if (success) {
if (key === 'EmailDomainWhitelist') {
value = value.split(',');
}
if (key === 'Price') {
value = parseFloat(value);
}
setInputs((inputs) => ({
...inputs,
[key]: value,
}));
} else {
showError(message);
}
setLoading(false);
};
const handleInputChange = async (e, { name, value }) => {
if (name === 'PasswordLoginEnabled' && inputs[name] === 'true') {
// block disabling password login
setShowPasswordWarningModal(true);
return;
}
if (
name === 'Notice' ||
(name.startsWith('SMTP') && name !== 'SMTPSSLEnabled') ||
name === 'ServerAddress' ||
name === 'WorkerUrl' ||
name === 'WorkerValidKey' ||
name === 'EpayId' ||
name === 'EpayKey' ||
name === 'Price' ||
name === 'PayAddress' ||
name === 'GitHubClientId' ||
name === 'GitHubClientSecret' ||
name === 'WeChatServerAddress' ||
name === 'WeChatServerToken' ||
name === 'WeChatAccountQRCodeImageURL' ||
name === 'TurnstileSiteKey' ||
name === 'TurnstileSecretKey' ||
name === 'EmailDomainWhitelist' ||
name === 'TopupGroupRatio' ||
name === 'TelegramBotToken' ||
name === 'TelegramBotName'
) {
setInputs((inputs) => ({ ...inputs, [name]: value }));
} else {
await updateOption(name, value);
}
};
const submitServerAddress = async () => {
let ServerAddress = removeTrailingSlash(inputs.ServerAddress);
await updateOption('ServerAddress', ServerAddress);
};
const submitWorker = async () => {
let WorkerUrl = removeTrailingSlash(inputs.WorkerUrl);
await updateOption('WorkerUrl', WorkerUrl);
if (inputs.WorkerValidKey !== '') {
await updateOption('WorkerValidKey', inputs.WorkerValidKey);
}
}
const submitPayAddress = async () => {
if (inputs.ServerAddress === '') {
showError('请先填写服务器地址');
return;
}
if (originInputs['TopupGroupRatio'] !== inputs.TopupGroupRatio) {
if (!verifyJSON(inputs.TopupGroupRatio)) {
showError('充值分组倍率不是合法的 JSON 字符串');
return;
}
await updateOption('TopupGroupRatio', inputs.TopupGroupRatio);
}
let PayAddress = removeTrailingSlash(inputs.PayAddress);
await updateOption('PayAddress', PayAddress);
if (inputs.EpayId !== '') {
await updateOption('EpayId', inputs.EpayId);
}
if (inputs.EpayKey !== undefined && inputs.EpayKey !== '') {
await updateOption('EpayKey', inputs.EpayKey);
}
await updateOption('Price', '' + inputs.Price);
};
const submitSMTP = async () => {
if (originInputs['SMTPServer'] !== inputs.SMTPServer) {
await updateOption('SMTPServer', inputs.SMTPServer);
}
if (originInputs['SMTPAccount'] !== inputs.SMTPAccount) {
await updateOption('SMTPAccount', inputs.SMTPAccount);
}
if (originInputs['SMTPFrom'] !== inputs.SMTPFrom) {
await updateOption('SMTPFrom', inputs.SMTPFrom);
}
if (
originInputs['SMTPPort'] !== inputs.SMTPPort &&
inputs.SMTPPort !== ''
) {
await updateOption('SMTPPort', inputs.SMTPPort);
}
if (
originInputs['SMTPToken'] !== inputs.SMTPToken &&
inputs.SMTPToken !== ''
) {
await updateOption('SMTPToken', inputs.SMTPToken);
}
};
const submitEmailDomainWhitelist = async () => {
if (
originInputs['EmailDomainWhitelist'] !==
inputs.EmailDomainWhitelist.join(',') &&
inputs.SMTPToken !== ''
) {
await updateOption(
'EmailDomainWhitelist',
inputs.EmailDomainWhitelist.join(','),
);
}
};
const submitWeChat = async () => {
if (originInputs['WeChatServerAddress'] !== inputs.WeChatServerAddress) {
await updateOption(
'WeChatServerAddress',
removeTrailingSlash(inputs.WeChatServerAddress),
);
}
if (
originInputs['WeChatAccountQRCodeImageURL'] !==
inputs.WeChatAccountQRCodeImageURL
) {
await updateOption(
'WeChatAccountQRCodeImageURL',
inputs.WeChatAccountQRCodeImageURL,
);
}
if (
originInputs['WeChatServerToken'] !== inputs.WeChatServerToken &&
inputs.WeChatServerToken !== ''
) {
await updateOption('WeChatServerToken', inputs.WeChatServerToken);
}
};
const submitGitHubOAuth = async () => {
if (originInputs['GitHubClientId'] !== inputs.GitHubClientId) {
await updateOption('GitHubClientId', inputs.GitHubClientId);
}
if (
originInputs['GitHubClientSecret'] !== inputs.GitHubClientSecret &&
inputs.GitHubClientSecret !== ''
) {
await updateOption('GitHubClientSecret', inputs.GitHubClientSecret);
}
};
const submitTelegramSettings = async () => {
// await updateOption('TelegramOAuthEnabled', inputs.TelegramOAuthEnabled);
await updateOption('TelegramBotToken', inputs.TelegramBotToken);
await updateOption('TelegramBotName', inputs.TelegramBotName);
};
const submitTurnstile = async () => {
if (originInputs['TurnstileSiteKey'] !== inputs.TurnstileSiteKey) {
await updateOption('TurnstileSiteKey', inputs.TurnstileSiteKey);
}
if (
originInputs['TurnstileSecretKey'] !== inputs.TurnstileSecretKey &&
inputs.TurnstileSecretKey !== ''
) {
await updateOption('TurnstileSecretKey', inputs.TurnstileSecretKey);
}
};
const submitNewRestrictedDomain = () => {
const localDomainList = inputs.EmailDomainWhitelist;
if (
restrictedDomainInput !== '' &&
!localDomainList.includes(restrictedDomainInput)
) {
setRestrictedDomainInput('');
setInputs({
...inputs,
EmailDomainWhitelist: [...localDomainList, restrictedDomainInput],
});
setEmailDomainWhitelist([
...EmailDomainWhitelist,
{
key: restrictedDomainInput,
text: restrictedDomainInput,
value: restrictedDomainInput,
},
]);
}
};
return (
<Grid columns={1}>
<Grid.Column>
<Form loading={loading} inverted={isDark}>
<Header as='h3' inverted={isDark}>
通用设置
</Header>
<Form.Group widths='equal'>
<Form.Input
label='服务器地址'
placeholder='例如：https://yourdomain.com'
value={inputs.ServerAddress}
name='ServerAddress'
onChange={handleInputChange}
/>
</Form.Group>
<Form.Button onClick={submitServerAddress}>
更新服务器地址
</Form.Button>
<Header as='h3' inverted={isDark}>
代理设置（支持 <a href='https://github.com/Calcium-Ion/new-api-worker' target='_blank' rel='noreferrer'>new-api-worker</a>）
</Header>
<Form.Group widths='equal'>
<Form.Input
label='Worker地址，不填写则不启用代理'
placeholder='例如：https://workername.yourdomain.workers.dev'
value={inputs.WorkerUrl}
name='WorkerUrl'
onChange={handleInputChange}
/>
<Form.Input
label='Worker密钥，根据你部署的 Worker 填写'
placeholder='例如：your_secret_key'
value={inputs.WorkerValidKey}
name='WorkerValidKey'
onChange={handleInputChange}
/>
</Form.Group>
<Form.Button onClick={submitWorker}>
更新Worker设置
</Form.Button>
<Divider />
<Header as='h3' inverted={isDark}>
支付设置（当前仅支持易支付接口，默认使用上方服务器地址作为回调地址！）
</Header>
<Form.Group widths='equal'>
<Form.Input
label='支付地址，不填写则不启用在线支付'
placeholder='例如：https://yourdomain.com'
value={inputs.PayAddress}
name='PayAddress'
onChange={handleInputChange}
/>
<Form.Input
label='易支付商户ID'
placeholder='例如：0001'
value={inputs.EpayId}
name='EpayId'
onChange={handleInputChange}
/>
<Form.Input
label='易支付商户密钥'
placeholder='敏感信息不会发送到前端显示'
value={inputs.EpayKey}
name='EpayKey'
onChange={handleInputChange}
/>
</Form.Group>
<Form.Group widths='equal'>
<Form.Input
label='回调地址，不填写则使用上方服务器地址作为回调地址'
placeholder='例如：https://yourdomain.com'
value={inputs.CustomCallbackAddress}
name='CustomCallbackAddress'
onChange={handleInputChange}
/>
<Form.Input
label='充值价格（x元/美金）'
placeholder='例如：7，就是7元/美金'
value={inputs.Price}
name='Price'
min={0}
onChange={handleInputChange}
/>
<Form.Input
label='最低充值美元数量（以美金为单位，如果使用额度请自行换算！）'
placeholder='例如：2，就是最低充值2$'
value={inputs.MinTopUp}
name='MinTopUp'
min={1}
onChange={handleInputChange}
/>
</Form.Group>
<Form.Group widths='equal'>
<Form.TextArea
label='充值分组倍率'
name='TopupGroupRatio'
onChange={handleInputChange}
style={{ minHeight: 250, fontFamily: 'JetBrains Mono, Consolas' }}
autoComplete='new-password'
value={inputs.TopupGroupRatio}
placeholder='为一个 JSON 文本，键为组名称，值为倍率'
/>
</Form.Group>
<Form.Button onClick={submitPayAddress}>更新支付设置</Form.Button>
<Divider />
<Header as='h3' inverted={isDark}>
配置登录注册
</Header>
<Form.Group inline>
<Form.Checkbox
checked={inputs.PasswordLoginEnabled === 'true'}
label='允许通过密码进行登录'
name='PasswordLoginEnabled'
onChange={handleInputChange}
/>
{showPasswordWarningModal && (
<Modal
open={showPasswordWarningModal}
onClose={() => setShowPasswordWarningModal(false)}
size={'tiny'}
style={{ maxWidth: '450px' }}
>
<Modal.Header>警告</Modal.Header>
<Modal.Content>
<p>
取消密码登录将导致所有未绑定其他登录方式的用户（包括管理员）无法通过密码登录，确认取消？
</p>
</Modal.Content>
<Modal.Actions>
<Button onClick={() => setShowPasswordWarningModal(false)}>
取消
</Button>
<Button
color='yellow'
onClick={async () => {
setShowPasswordWarningModal(false);
await updateOption('PasswordLoginEnabled', 'false');
}}
>
确定
</Button>
</Modal.Actions>
</Modal>
)}
<Form.Checkbox
checked={inputs.PasswordRegisterEnabled === 'true'}
label='允许通过密码进行注册'
name='PasswordRegisterEnabled'
onChange={handleInputChange}
/>
<Form.Checkbox
checked={inputs.EmailVerificationEnabled === 'true'}
label='通过密码注册时需要进行邮箱验证'
name='EmailVerificationEnabled'
onChange={handleInputChange}
/>
<Form.Checkbox
checked={inputs.GitHubOAuthEnabled === 'true'}
label='允许通过 GitHub 账户登录 & 注册'
name='GitHubOAuthEnabled'
onChange={handleInputChange}
/>
<Form.Checkbox
checked={inputs.WeChatAuthEnabled === 'true'}
label='允许通过微信登录 & 注册'
name='WeChatAuthEnabled'
onChange={handleInputChange}
/>
<Form.Checkbox
checked={inputs.TelegramOAuthEnabled === 'true'}
label='允许通过 Telegram 进行登录'
name='TelegramOAuthEnabled'
onChange={handleInputChange}
/>
</Form.Group>
<Form.Group inline>
<Form.Checkbox
checked={inputs.RegisterEnabled === 'true'}
label='允许新用户注册（此项为否时，新用户将无法以任何方式进行注册）'
name='RegisterEnabled'
onChange={handleInputChange}
/>
<Form.Checkbox
checked={inputs.TurnstileCheckEnabled === 'true'}
label='启用 Turnstile 用户校验'
name='TurnstileCheckEnabled'
onChange={handleInputChange}
/>
</Form.Group>
<Divider />
<Header as='h3' inverted={isDark}>
配置邮箱域名白名单
<Header.Subheader>
用以防止恶意用户利用临时邮箱批量注册
</Header.Subheader>
</Header>
<Form.Group widths={3}>
<Form.Checkbox
label='启用邮箱域名白名单'
name='EmailDomainRestrictionEnabled'
onChange={handleInputChange}
checked={inputs.EmailDomainRestrictionEnabled === 'true'}
/>
</Form.Group>
<Form.Group widths={3}>
<Form.Checkbox
label='启用邮箱别名限制（例如：ab.cd@gmail.com）'
name='EmailAliasRestrictionEnabled'
onChange={handleInputChange}
checked={inputs.EmailAliasRestrictionEnabled === 'true'}
/>
</Form.Group>
<Form.Group widths={2}>
<Form.Dropdown
label='允许的邮箱域名'
placeholder='允许的邮箱域名'
name='EmailDomainWhitelist'
required
fluid
multiple
selection
onChange={handleInputChange}
value={inputs.EmailDomainWhitelist}
autoComplete='new-password'
options={EmailDomainWhitelist}
/>
<Form.Input
label='添加新的允许的邮箱域名'
action={
<Button
type='button'
onClick={() => {
submitNewRestrictedDomain();
}}
>
填入
</Button>
}
onKeyDown={(e) => {
if (e.key === 'Enter') {
submitNewRestrictedDomain();
}
}}
autoComplete='new-password'
placeholder='输入新的允许的邮箱域名'
value={restrictedDomainInput}
onChange={(e, { value }) => {
setRestrictedDomainInput(value);
}}
/>
</Form.Group>
<Form.Button onClick={submitEmailDomainWhitelist}>
保存邮箱域名白名单设置
</Form.Button>
<Divider />
<Header as='h3' inverted={isDark}>
配置 SMTP
<Header.Subheader>用以支持系统的邮件发送</Header.Subheader>
</Header>
<Form.Group widths={3}>
<Form.Input
label='SMTP 服务器地址'
name='SMTPServer'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.SMTPServer}
placeholder='例如：smtp.qq.com'
/>
<Form.Input
label='SMTP 端口'
name='SMTPPort'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.SMTPPort}
placeholder='默认: 587'
/>
<Form.Input
label='SMTP 账户'
name='SMTPAccount'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.SMTPAccount}
placeholder='通常是邮箱地址'
/>
</Form.Group>
<Form.Group widths={3}>
<Form.Input
label='SMTP 发送者邮箱'
name='SMTPFrom'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.SMTPFrom}
placeholder='通常和邮箱地址保持一致'
/>
<Form.Input
label='SMTP 访问凭证'
name='SMTPToken'
onChange={handleInputChange}
type='password'
autoComplete='new-password'
checked={inputs.RegisterEnabled === 'true'}
placeholder='敏感信息不会发送到前端显示'
/>
</Form.Group>
<Form.Group widths={3}>
<Form.Checkbox
label='启用SMTP SSL（465端口强制开启）'
name='SMTPSSLEnabled'
onChange={handleInputChange}
checked={inputs.SMTPSSLEnabled === 'true'}
/>
</Form.Group>
<Form.Button onClick={submitSMTP}>保存 SMTP 设置</Form.Button>
<Divider />
<Header as='h3' inverted={isDark}>
配置 GitHub OAuth App
<Header.Subheader>
用以支持通过 GitHub 进行登录注册，
<a
href='https://github.com/settings/developers'
target='_blank'
rel='noreferrer'
>
点击此处
</a>
管理你的 GitHub OAuth App
</Header.Subheader>
</Header>
<Message>
Homepage URL 填 <code>{inputs.ServerAddress}</code>
，Authorization callback URL 填{' '}
<code>{`${inputs.ServerAddress}/oauth/github`}</code>
</Message>
<Form.Group widths={3}>
<Form.Input
label='GitHub Client ID'
name='GitHubClientId'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.GitHubClientId}
placeholder='输入你注册的 GitHub OAuth APP 的 ID'
/>
<Form.Input
label='GitHub Client Secret'
name='GitHubClientSecret'
onChange={handleInputChange}
type='password'
autoComplete='new-password'
value={inputs.GitHubClientSecret}
placeholder='敏感信息不会发送到前端显示'
/>
</Form.Group>
<Form.Button onClick={submitGitHubOAuth}>
保存 GitHub OAuth 设置
</Form.Button>
<Divider />
<Header as='h3' inverted={isDark}>
配置 WeChat Server
<Header.Subheader>
用以支持通过微信进行登录注册，
<a
href='https://github.com/songquanpeng/wechat-server'
target='_blank'
rel='noreferrer'
>
点击此处
</a>
了解 WeChat Server
</Header.Subheader>
</Header>
<Form.Group widths={3}>
<Form.Input
label='WeChat Server 服务器地址'
name='WeChatServerAddress'
placeholder='例如：https://yourdomain.com'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.WeChatServerAddress}
/>
<Form.Input
label='WeChat Server 访问凭证'
name='WeChatServerToken'
type='password'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.WeChatServerToken}
placeholder='敏感信息不会发送到前端显示'
/>
<Form.Input
label='微信公众号二维码图片链接'
name='WeChatAccountQRCodeImageURL'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.WeChatAccountQRCodeImageURL}
placeholder='输入一个图片链接'
/>
</Form.Group>
<Form.Button onClick={submitWeChat}>
保存 WeChat Server 设置
</Form.Button>
<Divider />
<Header as='h3' inverted={isDark}>
配置 Telegram 登录
</Header>
<Form.Group inline>
<Form.Input
label='Telegram Bot Token'
name='TelegramBotToken'
onChange={handleInputChange}
value={inputs.TelegramBotToken}
placeholder='输入你的 Telegram Bot Token'
/>
<Form.Input
label='Telegram Bot 名称'
name='TelegramBotName'
onChange={handleInputChange}
value={inputs.TelegramBotName}
placeholder='输入你的 Telegram Bot 名称'
/>
</Form.Group>
<Form.Button onClick={submitTelegramSettings}>
保存 Telegram 登录设置
</Form.Button>
<Divider />
<Header as='h3' inverted={isDark}>
配置 Turnstile
<Header.Subheader>
用以支持用户校验，
<a
href='https://dash.cloudflare.com/'
target='_blank'
rel='noreferrer'
>
点击此处
</a>
管理你的 Turnstile Sites，推荐选择 Invisible Widget Type
</Header.Subheader>
</Header>
<Form.Group widths={3}>
<Form.Input
label='Turnstile Site Key'
name='TurnstileSiteKey'
onChange={handleInputChange}
autoComplete='new-password'
value={inputs.TurnstileSiteKey}
placeholder='输入你注册的 Turnstile Site Key'
/>
<Form.Input
label='Turnstile Secret Key'
name='TurnstileSecretKey'
onChange={handleInputChange}
type='password'
autoComplete='new-password'
value={inputs.TurnstileSecretKey}
placeholder='敏感信息不会发送到前端显示'
/>
</Form.Group>
<Form.Button onClick={submitTurnstile}>
保存 Turnstile 设置
</Form.Button>
</Form>
</Grid.Column>
</Grid>
);
};
export default SystemSetting;
web/src/constants/channel.constants.js
......@@ -80,11 +80,12 @@ export const CHANNEL_OPTIONS = [
label: 'Google PaLM2'
},
{
value: 45,
value: 47,
color: 'blue',
label: '字节火山方舟、豆包、DeepSeek通用'
label: 'Xinference'
},
{ value: 25, color: 'green', label: 'Moonshot' },
{ value: 20, color: 'green', label: 'OpenRouter' },
{ value: 19, color: 'blue', label: '360 智脑' },
{ value: 23, color: 'teal', label: '腾讯混元' },
{ value: 31, color: 'green', label: '零一万物' },
......@@ -108,5 +109,10 @@ export const CHANNEL_OPTIONS = [
value: 44,
color: 'purple',
label: '嵌入模型：MokaAI M3E'
}
},
{
value: 45,
color: 'blue',
label: '字节火山方舟、豆包、DeepSeek通用'
},
];
web/src/helpers/render.js
......@@ -325,9 +325,8 @@ export function renderModelPrice(
return (
<>
<article>
<p>{i18next.t('提示价格：${{price}} = ${{total}} / 1M tokens', {
<p>{i18next.t('提示价格：${{price}} / 1M tokens', {
price: inputRatioPrice,
total: inputRatioPrice
})}</p>
<p>{i18next.t('补全价格：${{price}} * {{completionRatio}} = ${{total}} / 1M tokens (补全倍率: {{completionRatio}})', {
price: inputRatioPrice,
......@@ -445,9 +444,8 @@ export function renderAudioModelPrice(
return (
<>
<article>
<p>{i18next.t('提示价格：${{price}} = ${{total}} / 1M tokens', {
<p>{i18next.t('提示价格：${{price}} / 1M tokens', {
price: inputRatioPrice,
total: inputRatioPrice
})}</p>
<p>{i18next.t('补全价格：${{price}} * {{completionRatio}} = ${{total}} / 1M tokens (补全倍率: {{completionRatio}})', {
price: inputRatioPrice,
......@@ -654,3 +652,194 @@ export function stringToColor(str) {
let i = sum % colors.length;
return colors[i];
}
export function renderClaudeModelPrice(
inputTokens,
completionTokens,
modelRatio,
modelPrice = -1,
completionRatio,
groupRatio,
cacheTokens = 0,
cacheRatio = 1.0,
cacheCreationTokens = 0,
cacheCreationRatio = 1.0,
) {
const ratioLabel = false ? i18next.t('专属倍率') : i18next.t('分组倍率');
if (modelPrice !== -1) {
return i18next.t('模型价格：${{price}} * {{ratioType}}：{{ratio}} = ${{total}}', {
price: modelPrice,
ratioType: ratioLabel,
ratio: groupRatio,
total: modelPrice * groupRatio
});
} else {
if (completionRatio === undefined) {
completionRatio = 0;
}
const completionRatioValue = completionRatio || 0;
const inputRatioPrice = modelRatio * 2.0;
const completionRatioPrice = modelRatio * 2.0 * completionRatioValue;
let cacheRatioPrice = (modelRatio * 2.0 * cacheRatio).toFixed(2);
let cacheCreationRatioPrice = modelRatio * 2.0 * cacheCreationRatio;
// Calculate effective input tokens (non-cached + cached with ratio applied + cache creation with ratio applied)
const nonCachedTokens = inputTokens;
const effectiveInputTokens = nonCachedTokens +
(cacheTokens * cacheRatio) +
(cacheCreationTokens * cacheCreationRatio);
let price =
(effectiveInputTokens / 1000000) * inputRatioPrice * groupRatio +
(completionTokens / 1000000) * completionRatioPrice * groupRatio;
return (
<>
<article>
<p>{i18next.t('提示价格：${{price}} / 1M tokens', {
price: inputRatioPrice,
})}</p>
<p>{i18next.t('补全价格：${{price}} * {{ratio}} = ${{total}} / 1M tokens', {
price: inputRatioPrice,
ratio: completionRatio,
total: completionRatioPrice
})}</p>
{cacheTokens > 0 && (
<p>{i18next.t('缓存价格：${{price}} * {{ratio}} = ${{total}} / 1M tokens (缓存倍率: {{cacheRatio}})', {
price: inputRatioPrice,
ratio: cacheRatio,
total: cacheRatioPrice,
cacheRatio: cacheRatio
})}</p>
)}
{cacheCreationTokens > 0 && (
<p>{i18next.t('缓存创建价格：${{price}} * {{ratio}} = ${{total}} / 1M tokens (缓存创建倍率: {{cacheCreationRatio}})', {
price: inputRatioPrice,
ratio: cacheCreationRatio,
total: cacheCreationRatioPrice,
cacheCreationRatio: cacheCreationRatio
})}</p>
)}
<p></p>
<p>
{(cacheTokens > 0 || cacheCreationTokens > 0) ?
i18next.t('提示 {{nonCacheInput}} tokens / 1M tokens * ${{price}} + 缓存 {{cacheInput}} tokens / 1M tokens * ${{cachePrice}} + 缓存创建 {{cacheCreationInput}} tokens / 1M tokens * ${{cacheCreationPrice}} + 补全 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}', {
nonCacheInput: nonCachedTokens,
cacheInput: cacheTokens,
cacheRatio: cacheRatio,
cacheCreationInput: cacheCreationTokens,
cacheCreationRatio: cacheCreationRatio,
cachePrice: cacheRatioPrice,
cacheCreationPrice: cacheCreationRatioPrice,
price: inputRatioPrice,
completion: completionTokens,
compPrice: completionRatioPrice,
ratio: groupRatio,
total: price.toFixed(6)
}) :
i18next.t('提示 {{input}} tokens / 1M tokens * ${{price}} + 补全 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}', {
input: inputTokens,
price: inputRatioPrice,
completion: completionTokens,
compPrice: completionRatioPrice,
ratio: groupRatio,
total: price.toFixed(6)
})
}
</p>
<p>{i18next.t('仅供参考，以实际扣费为准')}</p>
</article>
</>
);
}
}
export function renderClaudeLogContent(
modelRatio,
completionRatio,
modelPrice = -1,
groupRatio,
cacheRatio = 1.0,
cacheCreationRatio = 1.0,
) {
const ratioLabel = false ? i18next.t('专属倍率') : i18next.t('分组倍率');
if (modelPrice !== -1) {
return i18next.t('模型价格 ${{price}}，{{ratioType}} {{ratio}}', {
price: modelPrice,
ratioType: ratioLabel,
ratio: groupRatio
});
} else {
return i18next.t('模型倍率 {{modelRatio}}，补全倍率 {{completionRatio}}，缓存倍率 {{cacheRatio}}，缓存创建倍率 {{cacheCreationRatio}}，{{ratioType}} {{ratio}}', {
modelRatio: modelRatio,
completionRatio: completionRatio,
cacheRatio: cacheRatio,
cacheCreationRatio: cacheCreationRatio,
ratioType: ratioLabel,
ratio: groupRatio
});
}
}
export function renderClaudeModelPriceSimple(
modelRatio,
modelPrice = -1,
groupRatio,
cacheTokens = 0,
cacheRatio = 1.0,
cacheCreationTokens = 0,
cacheCreationRatio = 1.0,
) {
const ratioLabel = false ? i18next.t('专属倍率') : i18next.t('分组');
if (modelPrice !== -1) {
return i18next.t('价格：${{price}} * {{ratioType}}：{{ratio}}', {
price: modelPrice,
ratioType: ratioLabel,
ratio: groupRatio
});
} else {
if (cacheTokens !== 0 || cacheCreationTokens !== 0) {
return i18next.t('模型: {{ratio}} * {{ratioType}}: {{groupRatio}} * 缓存: {{cacheRatio}}', {
ratio: modelRatio,
ratioType: ratioLabel,
groupRatio: groupRatio,
cacheRatio: cacheRatio,
cacheCreationRatio: cacheCreationRatio
});
} else {
return i18next.t('模型: {{ratio}} * {{ratioType}}: {{groupRatio}}', {
ratio: modelRatio,
ratioType: ratioLabel,
groupRatio: groupRatio
});
}
}
}
export function renderLogContent(
modelRatio,
completionRatio,
modelPrice = -1,
groupRatio
) {
const ratioLabel = false ? i18next.t('专属倍率') : i18next.t('分组倍率');
if (modelPrice !== -1) {
return i18next.t('模型价格 ${{price}}，{{ratioType}} {{ratio}}', {
price: modelPrice,
ratioType: ratioLabel,
ratio: groupRatio
});
} else {
return i18next.t('模型倍率 {{modelRatio}}，补全倍率 {{completionRatio}}，{{ratioType}} {{ratio}}', {
modelRatio: modelRatio,
completionRatio: completionRatio,
ratioType: ratioLabel,
ratio: groupRatio
});
}
}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or sign in to comment