Merge pull request #1100 from daggeryu/patch-4

fix aws claude-sonnet-4-20250514

README.en.md

@@ -1,10 +1,13 @@
+<p align="right">
+   <a href="./README.md">中文</a> | <strong>English</strong>
+</p>
 <div align="center">
 
 ![new-api](/web/public/logo.png)
 
 # New API
 
-🍥 Next Generation LLM Gateway and AI Asset Management System
+🍥 Next-Generation Large Model Gateway and AI Asset Management System
 
 <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>
 
@@ -33,171 +36,155 @@
 > This is an open-source project developed based on [One API](https://github.com/songquanpeng/one-api)
 
 > [!IMPORTANT]  
-> - Users must comply with OpenAI's [Terms of Use](https://openai.com/policies/terms-of-use) and relevant laws and regulations. Not to be used for illegal purposes.
-> - This project is for personal learning only. Stability is not guaranteed, and no technical support is provided.
+> - This project is for personal learning purposes only, with no guarantee of stability or technical support.
+> - Users must comply with OpenAI's [Terms of Use](https://openai.com/policies/terms-of-use) and **applicable laws and regulations**, and must not use it for illegal purposes.
+> - According to the [《Interim Measures for the Management of Generative Artificial Intelligence Services》](http://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm), please do not provide any unregistered generative AI services to the public in China.
+
+## 📚 Documentation
+
+For detailed documentation, please visit our official Wiki: [https://docs.newapi.pro/](https://docs.newapi.pro/)
 
 ## ✨ Key Features
 
-1. 🎨 New UI interface (some interfaces pending update)
-2. 🌍 Multi-language support (work in progress)
-3. 🎨 Added [Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy) interface support, [Integration Guide](Midjourney.md)
-4. 💰 Online recharge support, configurable in system settings:
-    - [x] EasyPay
-5. 🔍 Query usage quota by key:
-    - Works with [neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool)
-6. 📑 Configurable items per page in pagination
-7. 🔄 Compatible with original One API database (one-api.db)
-8. 💵 Support per-request model pricing, configurable in System Settings - Operation Settings
-9. ⚖️ Support channel **weighted random** selection
-10. 📈 Data dashboard (console)
-11. 🔒 Configurable model access per token
-12. 🤖 Telegram authorization login support:
-    1. System Settings - Configure Login Registration - Allow Telegram Login
-    2. Send /setdomain command to [@Botfather](https://t.me/botfather)
-    3. Select your bot, then enter http(s)://your-website/login
-    4. Telegram Bot name is the bot username without @
-13. 🎵 Added [Suno API](https://github.com/Suno-API/Suno-API) interface support, [Integration Guide](Suno.md)
-14. 🔄 Support for Rerank models, compatible with Cohere and Jina, can integrate with Dify, [Integration Guide](Rerank.md)
-15. ⚡ **[OpenAI Realtime API](https://platform.openai.com/docs/guides/realtime/integration)** - Support for OpenAI's Realtime API, including Azure channels
-16. 🧠 Support for setting reasoning effort through model name suffix:
-    - Add suffix `-high` to set high reasoning effort (e.g., `o3-mini-high`)
-    - Add suffix `-medium` to set medium reasoning effort
-    - Add suffix `-low` to set low reasoning effort
-17. 🔄 Thinking to content option `thinking_to_content` in `Channel->Edit->Channel Extra Settings`, default is `false`, when `true`, the `reasoning_content` of the thinking content will be converted to `<think>` tags and concatenated to the content returned.
-18. 🔄 Model rate limit, support setting total request limit and successful request limit in `System Settings->Rate Limit Settings`
-19. 💰 Cache billing support, when enabled can charge a configurable ratio for cache hits:
-    1. Set `Prompt Cache Ratio` in `System Settings -> Operation Settings`
-    2. Set `Prompt Cache Ratio` in channel settings, range 0-1 (e.g., 0.5 means 50% charge on cache hits)
+New API offers a wide range of features, please refer to [Features Introduction](https://docs.newapi.pro/wiki/features-introduction) for details:
+
+1. 🎨 Brand new UI interface
+2. 🌍 Multi-language support
+3. 💰 Online recharge functionality (YiPay)
+4. 🔍 Support for querying usage quotas with keys (works with [neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool))
+5. 🔄 Compatible with the original One API database
+6. 💵 Support for pay-per-use model pricing
+7. ⚖️ Support for weighted random channel selection
+8. 📈 Data dashboard (console)
+9. 🔒 Token grouping and model restrictions
+10. 🤖 Support for more authorization login methods (LinuxDO, Telegram, OIDC)
+11. 🔄 Support for Rerank models (Cohere and Jina), [API Documentation](https://docs.newapi.pro/api/jinaai-rerank)
+12. ⚡ Support for OpenAI Realtime API (including Azure channels), [API Documentation](https://docs.newapi.pro/api/openai-realtime)
+13. ⚡ Support for Claude Messages format, [API Documentation](https://docs.newapi.pro/api/anthropic-chat)
+14. Support for entering chat interface via /chat2link route
+15. 🧠 Support for setting reasoning effort through model name suffixes:
+    1. OpenAI o-series models
+        - Add `-high` suffix for high reasoning effort (e.g.: `o3-mini-high`)
+        - Add `-medium` suffix for medium reasoning effort (e.g.: `o3-mini-medium`)
+        - Add `-low` suffix for low reasoning effort (e.g.: `o3-mini-low`)
+    2. Claude thinking models
+        - Add `-thinking` suffix to enable thinking mode (e.g.: `claude-3-7-sonnet-20250219-thinking`)
+16. 🔄 Thinking-to-content functionality
+17. 🔄 Model rate limiting for users
+18. 💰 Cache billing support, which allows billing at a set ratio when cache is hit:
+    1. Set the `Prompt Cache Ratio` option in `System Settings-Operation Settings`
+    2. Set `Prompt Cache Ratio` in the channel, range 0-1, e.g., setting to 0.5 means billing at 50% when cache is hit
     3. Supported channels:
         - [x] OpenAI
-        - [x] Azure 
+        - [x] Azure
         - [x] DeepSeek
-        - [ ] Claude
+        - [x] Claude
 
 ## Model Support
-This version additionally supports:
-1. Third-party model **gpts** (gpt-4-gizmo-*)
-2. [Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy) interface, [Integration Guide](Midjourney.md)
-3. Custom channels with full API URL support
-4. [Suno API](https://github.com/Suno-API/Suno-API) interface, [Integration Guide](Suno.md)
-5. Rerank models, supporting [Cohere](https://cohere.ai/) and [Jina](https://jina.ai/), [Integration Guide](Rerank.md)
-6. Dify
 
-You can add custom models gpt-4-gizmo-* in channels. These are third-party models and cannot be called with official OpenAI keys.
+This version supports multiple models, please refer to [API Documentation-Relay Interface](https://docs.newapi.pro/api) for details:
 
-## Additional Configurations Beyond One API
-- `GENERATE_DEFAULT_TOKEN`: Generate initial token for new users, default `false`
-- `STREAMING_TIMEOUT`: Set streaming response timeout, default 60 seconds
-- `DIFY_DEBUG`: Output workflow and node info to client for Dify channel, default `true`
-- `FORCE_STREAM_OPTION`: Override client stream_options parameter, default `true`
-- `GET_MEDIA_TOKEN`: Calculate image tokens, default `true`
-- `GET_MEDIA_TOKEN_NOT_STREAM`: Calculate image tokens in non-stream mode, default `true`
-- `UPDATE_TASK`: Update async tasks (Midjourney, Suno), default `true`
-- `GEMINI_MODEL_MAP`: Specify Gemini model versions (v1/v1beta), format: "model:version", comma-separated
-- `COHERE_SAFETY_SETTING`: Cohere model [safety settings](https://docs.cohere.com/docs/safety-modes#overview), options: `NONE`, `CONTEXTUAL`, `STRICT`, default `NONE`
-- `GEMINI_VISION_MAX_IMAGE_NUM`: Gemini model maximum image number, default `16`, set to `-1` to disable
-- `MAX_FILE_DOWNLOAD_MB`: Maximum file download size in MB, default `20`
-- `CRYPTO_SECRET`: Encryption key for encrypting database content
-- `AZURE_DEFAULT_API_VERSION`: Azure channel default API version, if not specified in channel settings, use this version, default `2024-12-01-preview`
-- `NOTIFICATION_LIMIT_DURATION_MINUTE`: Duration of notification limit in minutes, default `10`
-- `NOTIFY_LIMIT_COUNT`: Maximum number of user notifications in the specified duration, default `2`
+1. Third-party models **gpts** (gpt-4-gizmo-*)
+2. Third-party channel [Midjourney-Proxy(Plus)](https://github.com/novicezk/midjourney-proxy) interface, [API Documentation](https://docs.newapi.pro/api/midjourney-proxy-image)
+3. Third-party channel [Suno API](https://github.com/Suno-API/Suno-API) interface, [API Documentation](https://docs.newapi.pro/api/suno-music)
+4. Custom channels, supporting full call address input
+5. Rerank models ([Cohere](https://cohere.ai/) and [Jina](https://jina.ai/)), [API Documentation](https://docs.newapi.pro/api/jinaai-rerank)
+6. Claude Messages format, [API Documentation](https://docs.newapi.pro/api/anthropic-chat)
+7. Dify, currently only supports chatflow
+
+## Environment Variable Configuration
+
+For detailed configuration instructions, please refer to [Installation Guide-Environment Variables Configuration](https://docs.newapi.pro/installation/environment-variables):
+
+- `GENERATE_DEFAULT_TOKEN`: Whether to generate initial tokens for newly registered users, default is `false`
+- `STREAMING_TIMEOUT`: Streaming response timeout, default is 60 seconds
+- `DIFY_DEBUG`: Whether to output workflow and node information for Dify channels, default is `true`
+- `FORCE_STREAM_OPTION`: Whether to override client stream_options parameter, default is `true`
+- `GET_MEDIA_TOKEN`: Whether to count image tokens, default is `true`
+- `GET_MEDIA_TOKEN_NOT_STREAM`: Whether to count image tokens in non-streaming cases, default is `true`
+- `UPDATE_TASK`: Whether to update asynchronous tasks (Midjourney, Suno), default is `true`
+- `COHERE_SAFETY_SETTING`: Cohere model safety settings, options are `NONE`, `CONTEXTUAL`, `STRICT`, default is `NONE`
+- `GEMINI_VISION_MAX_IMAGE_NUM`: Maximum number of images for Gemini models, default is `16`
+- `MAX_FILE_DOWNLOAD_MB`: Maximum file download size in MB, default is `20`
+- `CRYPTO_SECRET`: Encryption key used for encrypting database content
+- `AZURE_DEFAULT_API_VERSION`: Azure channel default API version, default is `2025-04-01-preview`
+- `NOTIFICATION_LIMIT_DURATION_MINUTE`: Notification limit duration, default is `10` minutes
+- `NOTIFY_LIMIT_COUNT`: Maximum number of user notifications within the specified duration, default is `2`
 
 ## Deployment
 
+For detailed deployment guides, please refer to [Installation Guide-Deployment Methods](https://docs.newapi.pro/installation):
+
 > [!TIP]
-> Latest Docker image: `calciumion/new-api:latest`  
-> Default account: root, password: 123456
+> Latest Docker image: `calciumion/new-api:latest`
 
-### Multi-Server Deployment
-- Must set `SESSION_SECRET` environment variable, otherwise login state will not be consistent across multiple servers.
-- If using a public Redis, must set `CRYPTO_SECRET` environment variable, otherwise Redis content will not be able to be obtained in multi-server deployment.
+### Multi-machine Deployment Considerations
+- Environment variable `SESSION_SECRET` must be set, otherwise login status will be inconsistent across multiple machines
+- If sharing Redis, `CRYPTO_SECRET` must be set, otherwise Redis content cannot be accessed across multiple machines
 
-### Requirements
-- Local database (default): SQLite (Docker deployment must mount `/data` directory)
-- Remote database: MySQL >= 5.7.8, PgSQL >= 9.6
+### Deployment Requirements
+- Local database (default): SQLite (Docker deployment must mount the `/data` directory)
+- Remote database: MySQL version >= 5.7.8, PgSQL version >= 9.6
 
-### Deployment with BT Panel
-Install BT Panel (**version 9.2.0** or above) from [BT Panel Official Website](https://www.bt.cn/new/download.html), choose the stable version script to download and install.  
-After installation, log in to BT Panel and click Docker in the menu bar. First-time access will prompt to install Docker service. Click Install Now and follow the prompts to complete installation.  
-After installation, find **New-API** in the app store, click install, configure basic options to complete installation.  
-[Pictorial Guide](BT.md)
+### Deployment Methods
 
-### Docker Deployment
+#### Using BaoTa Panel Docker Feature
+Install BaoTa Panel (version **9.2.0** or above), find **New-API** in the application store and install it.
+[Tutorial with images](./docs/BT.md)
 
-### Using Docker Compose (Recommended)
+#### Using Docker Compose (Recommended)
 ```shell
-# Clone project
+# Download the project
 git clone https://github.com/Calcium-Ion/new-api.git
 cd new-api
 # Edit docker-compose.yml as needed
-# nano docker-compose.yml
-# vim docker-compose.yml
 # Start
 docker-compose up -d
 ```
 
-#### Update Version
+#### Using Docker Image Directly
 ```shell
-docker-compose pull
-docker-compose up -d
-```
-
-### Direct Docker Image Usage
-```shell
-# SQLite deployment:
+# Using 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
 
-# MySQL deployment (add -e SQL_DSN="root:123456@tcp(localhost:3306)/oneapi"), modify database connection parameters as needed
-# Example:
+# Using 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
 ```
 
-#### Update Version
-```shell
-# Pull the latest image
-docker pull calciumion/new-api:latest
-# Stop and remove the old container
-docker stop new-api
-docker rm new-api
-# Run the new container with the same parameters as before
-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
-```
+## Channel Retry and Cache
+Channel retry functionality has been implemented, you can set the number of retries in `Settings->Operation Settings->General Settings`. It is **recommended to enable caching**.
 
-Alternatively, you can use Watchtower for automatic updates (not recommended, may cause database incompatibility):
-```shell
-docker run --rm -v /var/run/docker.sock:/var/run/docker.sock containrrr/watchtower -cR
-```
+### Cache Configuration Method
+1. `REDIS_CONN_STRING`: Set Redis as cache
+2. `MEMORY_CACHE_ENABLED`: Enable memory cache (no need to set manually if Redis is set)
 
-## Channel Retry
-Channel retry is implemented, configurable in `Settings->Operation Settings->General Settings`. **Cache recommended**.  
-If retry is enabled, the system will automatically use the next priority channel for the same request after a failed request.
+## API Documentation
 
-### Cache Configuration
-1. `REDIS_CONN_STRING`: Use Redis as cache
-    + Example: `REDIS_CONN_STRING=redis://default:redispw@localhost:49153`
-2. `MEMORY_CACHE_ENABLED`: Enable memory cache, default `false`
-    + Example: `MEMORY_CACHE_ENABLED=true`
+For detailed API documentation, please refer to [API Documentation](https://docs.newapi.pro/api):
 
-### Why Some Errors Don't Retry
-Error codes 400, 504, 524 won't retry
-### To Enable Retry for 400
-In `Channel->Edit`, set `Status Code Override` to:
-```json
-{
-  "400": "500"
-}
-```
-
-## Integration Guides
-- [Midjourney Integration](Midjourney.md)
-- [Suno Integration](Suno.md)
+- [Chat API](https://docs.newapi.pro/api/openai-chat)
+- [Image API](https://docs.newapi.pro/api/openai-image)
+- [Rerank API](https://docs.newapi.pro/api/jinaai-rerank)
+- [Realtime API](https://docs.newapi.pro/api/openai-realtime)
+- [Claude Chat API (messages)](https://docs.newapi.pro/api/anthropic-chat)
 
 ## Related Projects
 - [One API](https://github.com/songquanpeng/one-api): Original project
 - [Midjourney-Proxy](https://github.com/novicezk/midjourney-proxy): Midjourney interface support
-- [chatnio](https://github.com/Deeptrain-Community/chatnio): Next-gen AI B/C solution
-- [neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool): Query usage quota by key
+- [chatnio](https://github.com/Deeptrain-Community/chatnio): Next-generation AI one-stop B/C-end solution
+- [neko-api-key-tool](https://github.com/Calcium-Ion/neko-api-key-tool): Query usage quota with key
+
+Other projects based on New API:
+- [new-api-horizon](https://github.com/Calcium-Ion/new-api-horizon): High-performance optimized version of New API
+- [VoAPI](https://github.com/VoAPI/VoAPI): Frontend beautified version based on New API
+
+## Help and Support
+
+If you have any questions, please refer to [Help and Support](https://docs.newapi.pro/support):
+- [Community Interaction](https://docs.newapi.pro/support/community-interaction)
+- [Issue Feedback](https://docs.newapi.pro/support/feedback-issues)
+- [FAQ](https://docs.newapi.pro/support/faq)
 
 ## 🌟 Star History
 
-[![Star History Chart](https://api.star-history.com/svg?repos=Calcium-Ion/new-api&type=Date)](https://star-history.com/#Calcium-Ion/new-api&Date)
+[![Star History Chart](https://api.star-history.com/svg?repos=Calcium-Ion/new-api&type=Date)](https://star-history.com/#Calcium-Ion/new-api&Date)

README.md

@@ -107,7 +107,7 @@ New API提供了丰富的功能，详细特性请参考[特性说明](https://do
 - `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`
+- `AZURE_DEFAULT_API_VERSION`：Azure渠道默认API版本，默认 `2025-04-01-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`
 
@@ -130,7 +130,7 @@ New API提供了丰富的功能，详细特性请参考[特性说明](https://do
 
 #### 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)
+[图文教程](./docs/BT.md)
 
 #### 使用Docker Compose部署（推荐）
 ```shell

common/constants.go

@@ -62,6 +62,10 @@ var EmailDomainWhitelist = []string{
 	"yahoo.com",
 	"foxmail.com",
 }
+var EmailLoginAuthServerList = []string{
+	"smtp.sendcloud.net",
+	"smtp.azurecomm.net",
+}
 
 var DebugEnabled bool
 var MemoryCacheEnabled bool
@@ -236,6 +240,7 @@ const (
 	ChannelTypeBaiduV2        = 46
 	ChannelTypeXinference     = 47
 	ChannelTypeXai            = 48
+	ChannelTypeCoze           = 49
 	ChannelTypeDummy          // this one is only for count, do not add any channel after this
 
 )
@@ -290,4 +295,5 @@ var ChannelBaseURLs = []string{
 	"https://qianfan.baidubce.com",              //46
 	"",                                          //47
 	"https://api.x.ai",                          //48
+	"https://api.coze.cn",                       //49
 }

common/email.go

@@ -5,6 +5,7 @@ import (
 	"encoding/base64"
 	"fmt"
 	"net/smtp"
+	"slices"
 	"strings"
 	"time"
 )
@@ -79,7 +80,7 @@ func SendEmail(subject string, receiver string, content string) error {
 		if err != nil {
 			return err
 		}
-	} else if isOutlookServer(SMTPAccount) || SMTPServer == "smtp.azurecomm.net" {
+	} else if isOutlookServer(SMTPAccount) || slices.Contains(EmailLoginAuthServerList, SMTPServer) {
 		auth = LoginAuth(SMTPAccount, SMTPToken)
 		err = smtp.SendMail(addr, auth, SMTPFrom, to, mail)
 	} else {

common/limiter/limiter.go

@@ -0,0 +1,89 @@
+package limiter
+
+import (
+	"context"
+	_ "embed"
+	"fmt"
+	"github.com/go-redis/redis/v8"
+	"one-api/common"
+	"sync"
+)
+
+//go:embed lua/rate_limit.lua
+var rateLimitScript string
+
+type RedisLimiter struct {
+	client         *redis.Client
+	limitScriptSHA string
+}
+
+var (
+	instance *RedisLimiter
+	once     sync.Once
+)
+
+func New(ctx context.Context, r *redis.Client) *RedisLimiter {
+	once.Do(func() {
+		// 预加载脚本
+		limitSHA, err := r.ScriptLoad(ctx, rateLimitScript).Result()
+		if err != nil {
+			common.SysLog(fmt.Sprintf("Failed to load rate limit script: %v", err))
+		}
+		instance = &RedisLimiter{
+			client:         r,
+			limitScriptSHA: limitSHA,
+		}
+	})
+
+	return instance
+}
+
+func (rl *RedisLimiter) Allow(ctx context.Context, key string, opts ...Option) (bool, error) {
+	// 默认配置
+	config := &Config{
+		Capacity:  10,
+		Rate:      1,
+		Requested: 1,
+	}
+
+	// 应用选项模式
+	for _, opt := range opts {
+		opt(config)
+	}
+
+	// 执行限流
+	result, err := rl.client.EvalSha(
+		ctx,
+		rl.limitScriptSHA,
+		[]string{key},
+		config.Requested,
+		config.Rate,
+		config.Capacity,
+	).Int()
+
+	if err != nil {
+		return false, fmt.Errorf("rate limit failed: %w", err)
+	}
+	return result == 1, nil
+}
+
+// Config 配置选项模式
+type Config struct {
+	Capacity  int64
+	Rate      int64
+	Requested int64
+}
+
+type Option func(*Config)
+
+func WithCapacity(c int64) Option {
+	return func(cfg *Config) { cfg.Capacity = c }
+}
+
+func WithRate(r int64) Option {
+	return func(cfg *Config) { cfg.Rate = r }
+}
+
+func WithRequested(n int64) Option {
+	return func(cfg *Config) { cfg.Requested = n }
+}

common/limiter/lua/rate_limit.lua

@@ -0,0 +1,44 @@
+-- 令牌桶限流器
+-- KEYS[1]: 限流器唯一标识
+-- ARGV[1]: 请求令牌数 (通常为1)
+-- ARGV[2]: 令牌生成速率 (每秒)
+-- ARGV[3]: 桶容量
+
+local key = KEYS[1]
+local requested = tonumber(ARGV[1])
+local rate = tonumber(ARGV[2])
+local capacity = tonumber(ARGV[3])
+
+-- 获取当前时间（Redis服务器时间）
+local now = redis.call('TIME')
+local nowInSeconds = tonumber(now[1])
+
+-- 获取桶状态
+local bucket = redis.call('HMGET', key, 'tokens', 'last_time')
+local tokens = tonumber(bucket[1])
+local last_time = tonumber(bucket[2])
+
+-- 初始化桶（首次请求或过期）
+if not tokens or not last_time then
+    tokens = capacity
+    last_time = nowInSeconds
+else
+    -- 计算新增令牌
+    local elapsed = nowInSeconds - last_time
+    local add_tokens = elapsed * rate
+    tokens = math.min(capacity, tokens + add_tokens)
+    last_time = nowInSeconds
+end
+
+-- 判断是否允许请求
+local allowed = false
+if tokens >= requested then
+    tokens = tokens - requested
+    allowed = true
+end
+
+---- 更新桶状态并设置过期时间
+redis.call('HMSET', key, 'tokens', tokens, 'last_time', last_time)
+--redis.call('EXPIRE', key, math.ceil(capacity / rate) + 60) -- 适当延长过期时间
+
+return allowed and 1 or 0

common/utils.go

@@ -7,7 +7,6 @@ import (
 	"encoding/base64"
 	"encoding/json"
 	"fmt"
-	"github.com/pkg/errors"
 	"html/template"
 	"io"
 	"log"
@@ -22,6 +21,7 @@ import (
 	"time"
 
 	"github.com/google/uuid"
+	"github.com/pkg/errors"
 )
 
 func OpenBrowser(url string) {

constant/azure.go

@@ -0,0 +1,5 @@
+package constant
+
+import "time"
+
+var AzureNoRemoveDotTime = time.Date(2025, time.May, 10, 0, 0, 0, 0, time.UTC).Unix()

constant/env.go

@@ -16,6 +16,7 @@ var GeminiVisionMaxImageNum int
 var NotifyLimitCount int
 var NotificationLimitDurationMinute int
 var GenerateDefaultToken bool
+var ErrorLogEnabled bool
 
 //var GeminiModelMap = map[string]string{
 //	"gemini-1.0-pro": "v1",
@@ -30,12 +31,14 @@ func InitEnv() {
 	GetMediaToken = common.GetEnvOrDefaultBool("GET_MEDIA_TOKEN", true)
 	GetMediaTokenNotStream = common.GetEnvOrDefaultBool("GET_MEDIA_TOKEN_NOT_STREAM", true)
 	UpdateTask = common.GetEnvOrDefaultBool("UPDATE_TASK", true)
-	AzureDefaultAPIVersion = common.GetEnvOrDefaultString("AZURE_DEFAULT_API_VERSION", "2024-12-01-preview")
+	AzureDefaultAPIVersion = common.GetEnvOrDefaultString("AZURE_DEFAULT_API_VERSION", "2025-04-01-preview")
 	GeminiVisionMaxImageNum = common.GetEnvOrDefault("GEMINI_VISION_MAX_IMAGE_NUM", 16)
 	NotifyLimitCount = common.GetEnvOrDefault("NOTIFY_LIMIT_COUNT", 2)
 	NotificationLimitDurationMinute = common.GetEnvOrDefault("NOTIFICATION_LIMIT_DURATION_MINUTE", 10)
 	// GenerateDefaultToken 是否生成初始令牌，默认关闭。
 	GenerateDefaultToken = common.GetEnvOrDefaultBool("GENERATE_DEFAULT_TOKEN", false)
+	// 是否启用错误日志
+	ErrorLogEnabled = common.GetEnvOrDefaultBool("ERROR_LOG_ENABLED", false)
 
 	//modelVersionMapStr := strings.TrimSpace(os.Getenv("GEMINI_MODEL_MAP"))
 	//if modelVersionMapStr == "" {

controller/channel-billing.go

@@ -108,6 +108,13 @@ type DeepSeekUsageResponse struct {
 	} `json:"balance_infos"`
 }
 
+type OpenRouterCreditResponse struct {
+	Data struct {
+		TotalCredits float64 `json:"total_credits"`
+		TotalUsage   float64 `json:"total_usage"`
+	} `json:"data"`
+}
+
 // GetAuthHeader get auth header
 func GetAuthHeader(token string) http.Header {
 	h := http.Header{}
@@ -281,6 +288,22 @@ func updateChannelAIGC2DBalance(channel *model.Channel) (float64, error) {
 	return response.TotalAvailable, nil
 }
 
+func updateChannelOpenRouterBalance(channel *model.Channel) (float64, error) {
+	url := "https://openrouter.ai/api/v1/credits"
+	body, err := GetResponseBody("GET", url, channel, GetAuthHeader(channel.Key))
+	if err != nil {
+		return 0, err
+	}
+	response := OpenRouterCreditResponse{}
+	err = json.Unmarshal(body, &response)
+	if err != nil {
+		return 0, err
+	}
+	balance := response.Data.TotalCredits - response.Data.TotalUsage
+	channel.UpdateBalance(balance)
+	return balance, nil
+}
+
 func updateChannelBalance(channel *model.Channel) (float64, error) {
 	baseURL := common.ChannelBaseURLs[channel.Type]
 	if channel.GetBaseURL() == "" {
@@ -307,6 +330,8 @@ func updateChannelBalance(channel *model.Channel) (float64, error) {
 		return updateChannelSiliconFlowBalance(channel)
 	case common.ChannelTypeDeepSeek:
 		return updateChannelDeepSeekBalance(channel)
+	case common.ChannelTypeOpenRouter:
+		return updateChannelOpenRouterBalance(channel)
 	default:
 		return 0, errors.New("尚未实现")
 	}

controller/channel-test.go

@@ -103,7 +103,10 @@ func testChannel(channel *model.Channel, testModel string) (err error, openAIErr
 	}
 
 	request := buildTestRequest(testModel)
-	common.SysLog(fmt.Sprintf("testing channel %d with model %s , info %v ", channel.Id, testModel, info))
+	// 创建一个用于日志的 info 副本，移除 ApiKey
+	logInfo := *info
+	logInfo.ApiKey = ""
+	common.SysLog(fmt.Sprintf("testing channel %d with model %s , info %+v ", channel.Id, testModel, logInfo))
 
 	priceData, err := helper.ModelPriceHelper(c, info, 0, int(request.MaxTokens))
 	if err != nil {
@@ -186,7 +189,7 @@ func buildTestRequest(model string) *dto.GeneralOpenAIRequest {
 		return testRequest
 	}
 	// 并非Embedding 模型
-	if strings.HasPrefix(model, "o1") || strings.HasPrefix(model, "o3") {
+	if strings.HasPrefix(model, "o") {
 		testRequest.MaxCompletionTokens = 10
 	} else if strings.Contains(model, "thinking") {
 		if !strings.Contains(model, "claude") {

controller/log.go

@@ -196,7 +196,7 @@ func DeleteHistoryLogs(c *gin.Context) {
 		})
 		return
 	}
-	count, err := model.DeleteOldLog(targetTimestamp)
+	count, err := model.DeleteOldLog(c.Request.Context(), targetTimestamp, 100)
 	if err != nil {
 		c.JSON(http.StatusOK, gin.H{
 			"success": false,

controller/option.go

@@ -110,6 +110,15 @@ func UpdateOption(c *gin.Context) {
 			})
 			return
 		}
+	case "ModelRequestRateLimitGroup":
+		err = setting.CheckModelRequestRateLimitGroup(option.Value)
+		if err != nil {
+			c.JSON(http.StatusOK, gin.H{
+				"success": false,
+				"message": err.Error(),
+			})
+			return
+		}
 
 	}
 	err = model.UpdateOption(option.Key, option.Value)

controller/relay.go

@@ -4,12 +4,11 @@ import (
 	"bytes"
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
-	"github.com/gorilla/websocket"
 	"io"
 	"log"
 	"net/http"
 	"one-api/common"
+	constant2 "one-api/constant"
 	"one-api/dto"
 	"one-api/middleware"
 	"one-api/model"
@@ -19,12 +18,15 @@ import (
 	"one-api/relay/helper"
 	"one-api/service"
 	"strings"
+
+	"github.com/gin-gonic/gin"
+	"github.com/gorilla/websocket"
 )
 
 func relayHandler(c *gin.Context, relayMode int) *dto.OpenAIErrorWithStatusCode {
 	var err *dto.OpenAIErrorWithStatusCode
 	switch relayMode {
-	case relayconstant.RelayModeImagesGenerations:
+	case relayconstant.RelayModeImagesGenerations, relayconstant.RelayModeImagesEdits:
 		err = relay.ImageHelper(c)
 	case relayconstant.RelayModeAudioSpeech:
 		fallthrough
@@ -36,9 +38,31 @@ func relayHandler(c *gin.Context, relayMode int) *dto.OpenAIErrorWithStatusCode
 		err = relay.RerankHelper(c, relayMode)
 	case relayconstant.RelayModeEmbeddings:
 		err = relay.EmbeddingHelper(c)
+	case relayconstant.RelayModeResponses:
+		err = relay.ResponsesHelper(c)
 	default:
 		err = relay.TextHelper(c)
 	}
+
+	if constant2.ErrorLogEnabled && err != nil {
+		// 保存错误日志到mysql中
+		userId := c.GetInt("id")
+		tokenName := c.GetString("token_name")
+		modelName := c.GetString("original_model")
+		tokenId := c.GetInt("token_id")
+		userGroup := c.GetString("group")
+		channelId := c.GetInt("channel_id")
+		other := make(map[string]interface{})
+		other["error_type"] = err.Error.Type
+		other["error_code"] = err.Error.Code
+		other["status_code"] = err.StatusCode
+		other["channel_id"] = channelId
+		other["channel_name"] = c.GetString("channel_name")
+		other["channel_type"] = c.GetInt("channel_type")
+
+		model.RecordErrorLog(c, userId, channelId, modelName, tokenName, err.Error.Message, tokenId, 0, false, userGroup, other)
+	}
+
 	return err
 }
 

controller/user.go

@@ -592,7 +592,14 @@ func UpdateSelf(c *gin.Context) {
 		user.Password = "" // rollback to what it should be
 		cleanUser.Password = ""
 	}
-	updatePassword := user.Password != ""
+	updatePassword, err := checkUpdatePassword(user.OriginalPassword, user.Password, cleanUser.Id)
+	if err != nil {
+		c.JSON(http.StatusOK, gin.H{
+			"success": false,
+			"message": err.Error(),
+		})
+		return
+	}
 	if err := cleanUser.Update(updatePassword); err != nil {
 		c.JSON(http.StatusOK, gin.H{
 			"success": false,
@@ -608,6 +615,23 @@ func UpdateSelf(c *gin.Context) {
 	return
 }
 
+func checkUpdatePassword(originalPassword string, newPassword string, userId int) (updatePassword bool, err error) {
+	var currentUser *model.User
+	currentUser, err = model.GetUserById(userId, true)
+	if err != nil {
+		return
+	}
+	if !common.ValidatePasswordAndHash(originalPassword, currentUser.Password) {
+		err = fmt.Errorf("原密码错误")
+		return
+	}
+	if newPassword == "" {
+		return
+	}
+	updatePassword = true
+	return
+}
+
 func DeleteUser(c *gin.Context) {
 	id, err := strconv.Atoi(c.Param("id"))
 	if err != nil {

docker-compose.yml

@@ -15,6 +15,7 @@ services:
       - SQL_DSN=root:123456@tcp(mysql:3306)/new-api  # Point to the mysql service
       - REDIS_CONN_STRING=redis://redis
       - TZ=Asia/Shanghai
+      - ERROR_LOG_ENABLED=true # 是否启用错误日志记录
     #      - TIKTOKEN_CACHE_DIR=./tiktoken_cache  # 如果需要使用tiktoken_cache，请取消注释
     #      - SESSION_SECRET=random_string  # 多机部署时设置，必须修改这个随机字符串！！！！！！！
     #      - NODE_TYPE=slave  # Uncomment for slave node in multi-node deployment

docs/installation/BT.md

@@ -1,3 +1,3 @@
-密钥为环境变量SESSION_SECRET
-
-![8285bba413e770fe9620f1bf9b40d44e](https://github.com/user-attachments/assets/7a6fc03e-c457-45e4-b8f9-184508fc26b0)
+密钥为环境变量SESSION_SECRET
+
+![8285bba413e770fe9620f1bf9b40d44e](https://github.com/user-attachments/assets/7a6fc03e-c457-45e4-b8f9-184508fc26b0)

dto/claude.go

@@ -70,8 +70,9 @@ func (c *ClaudeMediaMessage) ParseMediaContent() []ClaudeMediaMessage {
 
 type ClaudeMessageSource struct {
 	Type      string `json:"type"`
-	MediaType string `json:"media_type"`
-	Data      any    `json:"data"`
+	MediaType string `json:"media_type,omitempty"`
+	Data      any    `json:"data,omitempty"`
+	Url       string `json:"url,omitempty"`
 }
 
 type ClaudeMessage struct {

dto/dalle.go

@@ -1,14 +1,20 @@
 package dto
 
+import "encoding/json"
+
 type ImageRequest struct {
-	Model          string `json:"model"`
-	Prompt         string `json:"prompt" binding:"required"`
-	N              int    `json:"n,omitempty"`
-	Size           string `json:"size,omitempty"`
-	Quality        string `json:"quality,omitempty"`
-	ResponseFormat string `json:"response_format,omitempty"`
-	Style          string `json:"style,omitempty"`
-	User           string `json:"user,omitempty"`
+	Model          string          `json:"model"`
+	Prompt         string          `json:"prompt" binding:"required"`
+	N              int             `json:"n,omitempty"`
+	Size           string          `json:"size,omitempty"`
+	Quality        string          `json:"quality,omitempty"`
+	ResponseFormat string          `json:"response_format,omitempty"`
+	Style          string          `json:"style,omitempty"`
+	User           string          `json:"user,omitempty"`
+	ExtraFields    json.RawMessage `json:"extra_fields,omitempty"`
+	Background     string          `json:"background,omitempty"`
+	Moderation     string          `json:"moderation,omitempty"`
+	OutputFormat   string          `json:"output_format,omitempty"`
 }
 
 type ImageResponse struct {

dto/openai_request.go

@@ -18,39 +18,41 @@ type FormatJsonSchema struct {
 }
 
 type GeneralOpenAIRequest struct {
-	Model               string            `json:"model,omitempty"`
-	Messages            []Message         `json:"messages,omitempty"`
-	Prompt              any               `json:"prompt,omitempty"`
-	Prefix              any               `json:"prefix,omitempty"`
-	Suffix              any               `json:"suffix,omitempty"`
-	Stream              bool              `json:"stream,omitempty"`
-	StreamOptions       *StreamOptions    `json:"stream_options,omitempty"`
-	MaxTokens           uint              `json:"max_tokens,omitempty"`
-	MaxCompletionTokens uint              `json:"max_completion_tokens,omitempty"`
-	ReasoningEffort     string            `json:"reasoning_effort,omitempty"`
-	Temperature         *float64          `json:"temperature,omitempty"`
-	TopP                float64           `json:"top_p,omitempty"`
-	TopK                int               `json:"top_k,omitempty"`
-	Stop                any               `json:"stop,omitempty"`
-	N                   int               `json:"n,omitempty"`
-	Input               any               `json:"input,omitempty"`
-	Instruction         string            `json:"instruction,omitempty"`
-	Size                string            `json:"size,omitempty"`
-	Functions           any               `json:"functions,omitempty"`
-	FrequencyPenalty    float64           `json:"frequency_penalty,omitempty"`
-	PresencePenalty     float64           `json:"presence_penalty,omitempty"`
-	ResponseFormat      *ResponseFormat   `json:"response_format,omitempty"`
-	EncodingFormat      any               `json:"encoding_format,omitempty"`
-	Seed                float64           `json:"seed,omitempty"`
-	Tools               []ToolCallRequest `json:"tools,omitempty"`
-	ToolChoice          any               `json:"tool_choice,omitempty"`
-	User                string            `json:"user,omitempty"`
-	LogProbs            bool              `json:"logprobs,omitempty"`
-	TopLogProbs         int               `json:"top_logprobs,omitempty"`
-	Dimensions          int               `json:"dimensions,omitempty"`
-	Modalities          any               `json:"modalities,omitempty"`
-	Audio               any               `json:"audio,omitempty"`
-	ExtraBody           any               `json:"extra_body,omitempty"`
+	Model               string         `json:"model,omitempty"`
+	Messages            []Message      `json:"messages,omitempty"`
+	Prompt              any            `json:"prompt,omitempty"`
+	Prefix              any            `json:"prefix,omitempty"`
+	Suffix              any            `json:"suffix,omitempty"`
+	Stream              bool           `json:"stream,omitempty"`
+	StreamOptions       *StreamOptions `json:"stream_options,omitempty"`
+	MaxTokens           uint           `json:"max_tokens,omitempty"`
+	MaxCompletionTokens uint           `json:"max_completion_tokens,omitempty"`
+	ReasoningEffort     string         `json:"reasoning_effort,omitempty"`
+	//Reasoning           json.RawMessage   `json:"reasoning,omitempty"`
+	Temperature      *float64          `json:"temperature,omitempty"`
+	TopP             float64           `json:"top_p,omitempty"`
+	TopK             int               `json:"top_k,omitempty"`
+	Stop             any               `json:"stop,omitempty"`
+	N                int               `json:"n,omitempty"`
+	Input            any               `json:"input,omitempty"`
+	Instruction      string            `json:"instruction,omitempty"`
+	Size             string            `json:"size,omitempty"`
+	Functions        any               `json:"functions,omitempty"`
+	FrequencyPenalty float64           `json:"frequency_penalty,omitempty"`
+	PresencePenalty  float64           `json:"presence_penalty,omitempty"`
+	ResponseFormat   *ResponseFormat   `json:"response_format,omitempty"`
+	EncodingFormat   any               `json:"encoding_format,omitempty"`
+	Seed             float64           `json:"seed,omitempty"`
+	Tools            []ToolCallRequest `json:"tools,omitempty"`
+	ToolChoice       any               `json:"tool_choice,omitempty"`
+	User             string            `json:"user,omitempty"`
+	LogProbs         bool              `json:"logprobs,omitempty"`
+	TopLogProbs      int               `json:"top_logprobs,omitempty"`
+	Dimensions       int               `json:"dimensions,omitempty"`
+	Modalities       any               `json:"modalities,omitempty"`
+	Audio            any               `json:"audio,omitempty"`
+	EnableThinking   any               `json:"enable_thinking,omitempty"` // ali
+	ExtraBody        any               `json:"extra_body,omitempty"`
 }
 
 type ToolCallRequest struct {
@@ -112,6 +114,7 @@ type MediaContent struct {
 	ImageUrl   any    `json:"image_url,omitempty"`
 	InputAudio any    `json:"input_audio,omitempty"`
 	File       any    `json:"file,omitempty"`
+	VideoUrl   any    `json:"video_url,omitempty"`
 }
 
 func (m *MediaContent) GetImageMedia() *MessageImageUrl {
@@ -156,11 +159,16 @@ type MessageFile struct {
 	FileId   string `json:"file_id,omitempty"`
 }
 
+type MessageVideoUrl struct {
+	Url string `json:"url"`
+}
+
 const (
 	ContentTypeText       = "text"
 	ContentTypeImageURL   = "image_url"
 	ContentTypeInputAudio = "input_audio"
 	ContentTypeFile       = "file"
+	ContentTypeVideoUrl   = "video_url" // 阿里百炼视频识别
 )
 
 func (m *Message) GetPrefix() bool {
@@ -344,6 +352,15 @@ func (m *Message) ParseContent() []MediaContent {
 						}
 					}
 				}
+			case ContentTypeVideoUrl:
+				if videoUrl, ok := contentItem["video_url"].(string); ok {
+					contentList = append(contentList, MediaContent{
+						Type: ContentTypeVideoUrl,
+						VideoUrl: &MessageVideoUrl{
+							Url: videoUrl,
+						},
+					})
+				}
 			}
 		}
 	}
@@ -353,3 +370,49 @@ func (m *Message) ParseContent() []MediaContent {
 	}
 	return contentList
 }
+
+type OpenAIResponsesRequest struct {
+	Model              string               `json:"model"`
+	Input              json.RawMessage      `json:"input,omitempty"`
+	Include            json.RawMessage      `json:"include,omitempty"`
+	Instructions       json.RawMessage      `json:"instructions,omitempty"`
+	MaxOutputTokens    uint                 `json:"max_output_tokens,omitempty"`
+	Metadata           json.RawMessage      `json:"metadata,omitempty"`
+	ParallelToolCalls  bool                 `json:"parallel_tool_calls,omitempty"`
+	PreviousResponseID string               `json:"previous_response_id,omitempty"`
+	Reasoning          *Reasoning           `json:"reasoning,omitempty"`
+	ServiceTier        string               `json:"service_tier,omitempty"`
+	Store              bool                 `json:"store,omitempty"`
+	Stream             bool                 `json:"stream,omitempty"`
+	Temperature        float64              `json:"temperature,omitempty"`
+	Text               json.RawMessage      `json:"text,omitempty"`
+	ToolChoice         json.RawMessage      `json:"tool_choice,omitempty"`
+	Tools              []ResponsesToolsCall `json:"tools,omitempty"`
+	TopP               float64              `json:"top_p,omitempty"`
+	Truncation         string               `json:"truncation,omitempty"`
+	User               string               `json:"user,omitempty"`
+}
+
+type Reasoning struct {
+	Effort  string `json:"effort,omitempty"`
+	Summary string `json:"summary,omitempty"`
+}
+
+type ResponsesToolsCall struct {
+	Type string `json:"type"`
+	// Web Search
+	UserLocation      json.RawMessage `json:"user_location,omitempty"`
+	SearchContextSize string          `json:"search_context_size,omitempty"`
+	// File Search
+	VectorStoreIds []string        `json:"vector_store_ids,omitempty"`
+	MaxNumResults  uint            `json:"max_num_results,omitempty"`
+	Filters        json.RawMessage `json:"filters,omitempty"`
+	// Computer Use
+	DisplayWidth  uint   `json:"display_width,omitempty"`
+	DisplayHeight uint   `json:"display_height,omitempty"`
+	Environment   string `json:"environment,omitempty"`
+	// Function
+	Name        string          `json:"name,omitempty"`
+	Description string          `json:"description,omitempty"`
+	Parameters  json.RawMessage `json:"parameters,omitempty"`
+}

dto/openai_response.go

@@ -1,5 +1,7 @@
 package dto
 
+import "encoding/json"
+
 type SimpleResponse struct {
 	Usage `json:"usage"`
 	Error *OpenAIError `json:"error"`
@@ -166,10 +168,93 @@ type CompletionsStreamResponse struct {
 }
 
 type Usage struct {
-	PromptTokens           int                `json:"prompt_tokens"`
-	CompletionTokens       int                `json:"completion_tokens"`
-	TotalTokens            int                `json:"total_tokens"`
-	PromptCacheHitTokens   int                `json:"prompt_cache_hit_tokens,omitempty"`
+	PromptTokens         int `json:"prompt_tokens"`
+	CompletionTokens     int `json:"completion_tokens"`
+	TotalTokens          int `json:"total_tokens"`
+	PromptCacheHitTokens int `json:"prompt_cache_hit_tokens,omitempty"`
+
 	PromptTokensDetails    InputTokenDetails  `json:"prompt_tokens_details"`
 	CompletionTokenDetails OutputTokenDetails `json:"completion_tokens_details"`
+	InputTokens            int                `json:"input_tokens"`
+	OutputTokens           int                `json:"output_tokens"`
+	InputTokensDetails     *InputTokenDetails `json:"input_tokens_details"`
+}
+
+type InputTokenDetails struct {
+	CachedTokens         int `json:"cached_tokens"`
+	CachedCreationTokens int `json:"-"`
+	TextTokens           int `json:"text_tokens"`
+	AudioTokens          int `json:"audio_tokens"`
+	ImageTokens          int `json:"image_tokens"`
+}
+
+type OutputTokenDetails struct {
+	TextTokens      int `json:"text_tokens"`
+	AudioTokens     int `json:"audio_tokens"`
+	ReasoningTokens int `json:"reasoning_tokens"`
+}
+
+type OpenAIResponsesResponse struct {
+	ID                 string               `json:"id"`
+	Object             string               `json:"object"`
+	CreatedAt          int                  `json:"created_at"`
+	Status             string               `json:"status"`
+	Error              *OpenAIError         `json:"error,omitempty"`
+	IncompleteDetails  *IncompleteDetails   `json:"incomplete_details,omitempty"`
+	Instructions       string               `json:"instructions"`
+	MaxOutputTokens    int                  `json:"max_output_tokens"`
+	Model              string               `json:"model"`
+	Output             []ResponsesOutput    `json:"output"`
+	ParallelToolCalls  bool                 `json:"parallel_tool_calls"`
+	PreviousResponseID string               `json:"previous_response_id"`
+	Reasoning          *Reasoning           `json:"reasoning"`
+	Store              bool                 `json:"store"`
+	Temperature        float64              `json:"temperature"`
+	ToolChoice         string               `json:"tool_choice"`
+	Tools              []ResponsesToolsCall `json:"tools"`
+	TopP               float64              `json:"top_p"`
+	Truncation         string               `json:"truncation"`
+	Usage              *Usage               `json:"usage"`
+	User               json.RawMessage      `json:"user"`
+	Metadata           json.RawMessage      `json:"metadata"`
+}
+
+type IncompleteDetails struct {
+	Reasoning string `json:"reasoning"`
+}
+
+type ResponsesOutput struct {
+	Type    string                   `json:"type"`
+	ID      string                   `json:"id"`
+	Status  string                   `json:"status"`
+	Role    string                   `json:"role"`
+	Content []ResponsesOutputContent `json:"content"`
+}
+
+type ResponsesOutputContent struct {
+	Type        string        `json:"type"`
+	Text        string        `json:"text"`
+	Annotations []interface{} `json:"annotations"`
+}
+
+const (
+	BuildInToolWebSearchPreview = "web_search_preview"
+	BuildInToolFileSearch       = "file_search"
+)
+
+const (
+	BuildInCallWebSearchCall = "web_search_call"
+)
+
+const (
+	ResponsesOutputTypeItemAdded = "response.output_item.added"
+	ResponsesOutputTypeItemDone  = "response.output_item.done"
+)
+
+// ResponsesStreamResponse 用于处理 /v1/responses 流式响应
+type ResponsesStreamResponse struct {
+	Type     string                   `json:"type"`
+	Response *OpenAIResponsesResponse `json:"response,omitempty"`
+	Delta    string                   `json:"delta,omitempty"`
+	Item     *ResponsesOutput         `json:"item,omitempty"`
 }

dto/realtime.go

@@ -43,20 +43,6 @@ type RealtimeUsage struct {
 	OutputTokenDetails OutputTokenDetails `json:"output_token_details"`
 }
 
-type InputTokenDetails struct {
-	CachedTokens         int `json:"cached_tokens"`
-	CachedCreationTokens int `json:"-"`
-	TextTokens           int `json:"text_tokens"`
-	AudioTokens          int `json:"audio_tokens"`
-	ImageTokens          int `json:"image_tokens"`
-}
-
-type OutputTokenDetails struct {
-	TextTokens      int `json:"text_tokens"`
-	AudioTokens     int `json:"audio_tokens"`
-	ReasoningTokens int `json:"reasoning_tokens"`
-}
-
 type RealtimeSession struct {
 	Modalities              []string                `json:"modalities"`
 	Instructions            string                  `json:"instructions"`

main.go

@@ -74,12 +74,14 @@ func main() {
 	}
 
 	// Initialize model settings
-	operation_setting.InitModelSettings()
+	operation_setting.InitRatioSettings()
 	// Initialize constants
 	constant.InitEnv()
 	// Initialize options
 	model.InitOptionMap()
 
+	service.InitTokenEncoders()
+
 	if common.RedisEnabled {
 		// for compatibility with old versions
 		common.MemoryCacheEnabled = true
@@ -87,9 +89,22 @@ func main() {
 	if common.MemoryCacheEnabled {
 		common.SysLog("memory cache enabled")
 		common.SysError(fmt.Sprintf("sync frequency: %d seconds", common.SyncFrequency))
-		model.InitChannelCache()
-	}
-	if common.MemoryCacheEnabled {
+
+		// Add panic recovery and retry for InitChannelCache
+		func() {
+			defer func() {
+				if r := recover(); r != nil {
+					common.SysError(fmt.Sprintf("InitChannelCache panic: %v, retrying once", r))
+					// Retry once
+					_, fixErr := model.FixAbility()
+					if fixErr != nil {
+						common.SysError(fmt.Sprintf("InitChannelCache failed: %s", fixErr.Error()))
+					}
+				}
+			}()
+			model.InitChannelCache()
+		}()
+
 		go model.SyncOptions(common.SyncFrequency)
 		go model.SyncChannelCache(common.SyncFrequency)
 	}
@@ -133,8 +148,6 @@ func main() {
 		common.SysLog("pprof enabled")
 	}
 
-	service.InitTokenEncoders()
-
 	// Initialize HTTP server
 	server := gin.New()
 	server.Use(gin.CustomRecovery(func(c *gin.Context, err any) {

middleware/distributor.go

@@ -162,7 +162,7 @@ func getModelRequest(c *gin.Context) (*ModelRequest, bool, error) {
 		}
 		c.Set("platform", string(constant.TaskPlatformSuno))
 		c.Set("relay_mode", relayMode)
-	} else if !strings.HasPrefix(c.Request.URL.Path, "/v1/audio/transcriptions") {
+	} else if !strings.HasPrefix(c.Request.URL.Path, "/v1/audio/transcriptions") && !strings.HasPrefix(c.Request.URL.Path, "/v1/images/edits") {
 		err = common.UnmarshalBodyReusable(c, &modelRequest)
 	}
 	if err != nil {
@@ -184,6 +184,8 @@ func getModelRequest(c *gin.Context) (*ModelRequest, bool, error) {
 	}
 	if strings.HasPrefix(c.Request.URL.Path, "/v1/images/generations") {
 		modelRequest.Model = common.GetStringIfEmpty(modelRequest.Model, "dall-e")
+	} else if strings.HasPrefix(c.Request.URL.Path, "/v1/images/edits") {
+		modelRequest.Model = common.GetStringIfEmpty(c.PostForm("model"), "gpt-image-1")
 	}
 	if strings.HasPrefix(c.Request.URL.Path, "/v1/audio") {
 		relayMode := relayconstant.RelayModeAudioSpeech
@@ -211,6 +213,7 @@ func SetupContextForSelectedChannel(c *gin.Context, channel *model.Channel, mode
 	c.Set("channel_id", channel.Id)
 	c.Set("channel_name", channel.Name)
 	c.Set("channel_type", channel.Type)
+	c.Set("channel_create_time", channel.CreatedTime)
 	c.Set("channel_setting", channel.GetSetting())
 	c.Set("param_override", channel.GetParamOverride())
 	if nil != channel.OpenAIOrganization && "" != *channel.OpenAIOrganization {
@@ -237,5 +240,7 @@ func SetupContextForSelectedChannel(c *gin.Context, channel *model.Channel, mode
 		c.Set("api_version", channel.Other)
 	case common.ChannelTypeMokaAI:
 		c.Set("api_version", channel.Other)
+	case common.ChannelTypeCoze:
+		c.Set("bot_id", channel.Other)
 	}
 }

middleware/model-rate-limit.go

@@ -5,6 +5,8 @@ import (
 	"fmt"
 	"net/http"
 	"one-api/common"
+	"one-api/common/limiter"
+	"one-api/constant"
 	"one-api/setting"
 	"strconv"
 	"time"
@@ -78,21 +80,9 @@ func redisRateLimitHandler(duration int64, totalMaxCount, successMaxCount int) g
 		ctx := context.Background()
 		rdb := common.RDB
 
-		// 1. 检查总请求数限制（当totalMaxCount为0时会自动跳过）
-		totalKey := fmt.Sprintf("rateLimit:%s:%s", ModelRequestRateLimitCountMark, userId)
-		allowed, err := checkRedisRateLimit(ctx, rdb, totalKey, totalMaxCount, duration)
-		if err != nil {
-			fmt.Println("检查总请求数限制失败:", err.Error())
-			abortWithOpenAiMessage(c, http.StatusInternalServerError, "rate_limit_check_failed")
-			return
-		}
-		if !allowed {
-			abortWithOpenAiMessage(c, http.StatusTooManyRequests, fmt.Sprintf("您已达到总请求数限制：%d分钟内最多请求%d次，包括失败次数，请检查您的请求是否正确", setting.ModelRequestRateLimitDurationMinutes, totalMaxCount))
-		}
-
-		// 2. 检查成功请求数限制
+		// 1. 检查成功请求数限制
 		successKey := fmt.Sprintf("rateLimit:%s:%s", ModelRequestRateLimitSuccessCountMark, userId)
-		allowed, err = checkRedisRateLimit(ctx, rdb, successKey, successMaxCount, duration)
+		allowed, err := checkRedisRateLimit(ctx, rdb, successKey, successMaxCount, duration)
 		if err != nil {
 			fmt.Println("检查成功请求数限制失败:", err.Error())
 			abortWithOpenAiMessage(c, http.StatusInternalServerError, "rate_limit_check_failed")
@@ -103,8 +93,29 @@ func redisRateLimitHandler(duration int64, totalMaxCount, successMaxCount int) g
 			return
 		}
 
-		// 3. 记录总请求（当totalMaxCount为0时会自动跳过）
-		recordRedisRequest(ctx, rdb, totalKey, totalMaxCount)
+		//2.检查总请求数限制并记录总请求（当totalMaxCount为0时会自动跳过，使用令牌桶限流器
+		if totalMaxCount > 0 {
+			totalKey := fmt.Sprintf("rateLimit:%s", userId)
+			// 初始化
+			tb := limiter.New(ctx, rdb)
+			allowed, err = tb.Allow(
+				ctx,
+				totalKey,
+				limiter.WithCapacity(int64(totalMaxCount)*duration),
+				limiter.WithRate(int64(totalMaxCount)),
+				limiter.WithRequested(duration),
+			)
+
+			if err != nil {
+				fmt.Println("检查总请求数限制失败:", err.Error())
+				abortWithOpenAiMessage(c, http.StatusInternalServerError, "rate_limit_check_failed")
+				return
+			}
+
+			if !allowed {
+				abortWithOpenAiMessage(c, http.StatusTooManyRequests, fmt.Sprintf("您已达到总请求数限制：%d分钟内最多请求%d次，包括失败次数，请检查您的请求是否正确", setting.ModelRequestRateLimitDurationMinutes, totalMaxCount))
+			}
+		}
 
 		// 4. 处理请求
 		c.Next()
@@ -165,6 +176,19 @@ func ModelRequestRateLimit() func(c *gin.Context) {
 		totalMaxCount := setting.ModelRequestRateLimitCount
 		successMaxCount := setting.ModelRequestRateLimitSuccessCount
 
+		// 获取分组
+		group := c.GetString("token_group")
+		if group == "" {
+			group = c.GetString(constant.ContextKeyUserGroup)
+		}
+
+		//获取分组的限流配置
+		groupTotalCount, groupSuccessCount, found := setting.GetGroupRateLimit(group)
+		if found {
+			totalMaxCount = groupTotalCount
+			successMaxCount = groupSuccessCount
+		}
+
 		// 根据存储类型选择并执行限流处理器
 		if common.RedisEnabled {
 			redisRateLimitHandler(duration, totalMaxCount, successMaxCount)(c)

model/ability.go

@@ -50,7 +50,7 @@ func getPriority(group string, model string, retry int) (int, error) {
 	err := DB.Model(&Ability{}).
 		Select("DISTINCT(priority)").
 		Where(groupCol+" = ? and model = ? and enabled = "+trueVal, group, model).
-		Order("priority DESC").              // 按优先级降序排序
+		Order("priority DESC"). // 按优先级降序排序
 		Pluck("priority", &priorities).Error // Pluck用于将查询的结果直接扫描到一个切片中
 
 	if err != nil {
@@ -261,12 +261,28 @@ func FixAbility() (int, error) {
 		common.SysError(fmt.Sprintf("Get channel ids from channel table failed: %s", err.Error()))
 		return 0, err
 	}
-	// Delete abilities of channels that are not in channel table
-	err = DB.Where("channel_id NOT IN (?)", channelIds).Delete(&Ability{}).Error
-	if err != nil {
-		common.SysError(fmt.Sprintf("Delete abilities of channels that are not in channel table failed: %s", err.Error()))
-		return 0, err
+
+	// Delete abilities of channels that are not in channel table - in batches to avoid too many placeholders
+	if len(channelIds) > 0 {
+		// Process deletion in chunks to avoid "too many placeholders" error
+		for _, chunk := range lo.Chunk(channelIds, 100) {
+			err = DB.Where("channel_id NOT IN (?)", chunk).Delete(&Ability{}).Error
+			if err != nil {
+				common.SysError(fmt.Sprintf("Delete abilities of channels (batch) that are not in channel table failed: %s", err.Error()))
+				return 0, err
+			}
+		}
+	} else {
+		// If no channels exist, delete all abilities
+		err = DB.Delete(&Ability{}).Error
+		if err != nil {
+			common.SysError(fmt.Sprintf("Delete all abilities failed: %s", err.Error()))
+			return 0, err
+		}
+		common.SysLog("Delete all abilities successfully")
+		return 0, nil
 	}
+
 	common.SysLog(fmt.Sprintf("Delete abilities of channels that are not in channel table successfully, ids: %v", channelIds))
 	count += len(channelIds)
 
@@ -275,17 +291,26 @@ func FixAbility() (int, error) {
 	err = DB.Table("abilities").Distinct("channel_id").Pluck("channel_id", &abilityChannelIds).Error
 	if err != nil {
 		common.SysError(fmt.Sprintf("Get channel ids from abilities table failed: %s", err.Error()))
-		return 0, err
+		return count, err
 	}
+
 	var channels []Channel
 	if len(abilityChannelIds) == 0 {
 		err = DB.Find(&channels).Error
 	} else {
-		err = DB.Where("id NOT IN (?)", abilityChannelIds).Find(&channels).Error
-	}
-	if err != nil {
-		return 0, err
+		// Process query in chunks to avoid "too many placeholders" error
+		err = nil
+		for _, chunk := range lo.Chunk(abilityChannelIds, 100) {
+			var channelsChunk []Channel
+			err = DB.Where("id NOT IN (?)", chunk).Find(&channelsChunk).Error
+			if err != nil {
+				common.SysError(fmt.Sprintf("Find channels not in abilities table failed: %s", err.Error()))
+				return count, err
+			}
+			channels = append(channels, channelsChunk...)
+		}
 	}
+
 	for _, channel := range channels {
 		err := channel.UpdateAbilities(nil)
 		if err != nil {

model/cache.go

@@ -16,6 +16,9 @@ var channelsIDM map[int]*Channel
 var channelSyncLock sync.RWMutex
 
 func InitChannelCache() {
+	if !common.MemoryCacheEnabled {
+		return
+	}
 	newChannelId2channel := make(map[int]*Channel)
 	var channels []*Channel
 	DB.Where("status = ?", common.ChannelStatusEnabled).Find(&channels)
@@ -84,9 +87,11 @@ func CacheGetRandomSatisfiedChannel(group string, model string, retry int) (*Cha
 	if !common.MemoryCacheEnabled {
 		return GetRandomSatisfiedChannel(group, model, retry)
 	}
+
 	channelSyncLock.RLock()
-	defer channelSyncLock.RUnlock()
 	channels := group2model2channels[group][model]
+	channelSyncLock.RUnlock()
+
 	if len(channels) == 0 {
 		return nil, errors.New("channel not found")
 	}

model/channel.go

@@ -46,6 +46,17 @@ func (channel *Channel) GetModels() []string {
 	return strings.Split(strings.Trim(channel.Models, ","), ",")
 }
 
+func (channel *Channel) GetGroups() []string {
+	if channel.Group == "" {
+		return []string{}
+	}
+	groups := strings.Split(strings.Trim(channel.Group, ","), ",")
+	for i, group := range groups {
+		groups[i] = strings.TrimSpace(group)
+	}
+	return groups
+}
+
 func (channel *Channel) GetOtherInfo() map[string]interface{} {
 	otherInfo := make(map[string]interface{})
 	if channel.OtherInfo != "" {
@@ -119,10 +130,15 @@ func SearchChannels(keyword string, group string, model string, idSort bool) ([]
 
 	// 如果是 PostgreSQL，使用双引号
 	if common.UsingPostgreSQL {
-		keyCol = `"key"`
 		modelsCol = `"models"`
 	}
 
+	baseURLCol := "`base_url`"
+	// 如果是 PostgreSQL，使用双引号
+	if common.UsingPostgreSQL {
+		baseURLCol = `"base_url"`
+	}
+
 	order := "priority desc"
 	if idSort {
 		order = "id desc"
@@ -142,11 +158,11 @@ func SearchChannels(keyword string, group string, model string, idSort bool) ([]
 			// sqlite, PostgreSQL
 			groupCondition = `(',' || ` + groupCol + ` || ',') LIKE ?`
 		}
-		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ?) AND " + modelsCol + ` LIKE ? AND ` + groupCondition
-		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+model+"%", "%,"+group+",%")
+		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ? OR " + baseURLCol + " LIKE ?) AND " + modelsCol + ` LIKE ? AND ` + groupCondition
+		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+keyword+"%", "%"+model+"%", "%,"+group+",%")
 	} else {
-		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ?) AND " + modelsCol + " LIKE ?"
-		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+model+"%")
+		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ? OR " + baseURLCol + " LIKE ?) AND " + modelsCol + " LIKE ?"
+		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+keyword+"%", "%"+model+"%")
 	}
 
 	// 执行查询
@@ -450,6 +466,12 @@ func SearchTags(keyword string, group string, model string, idSort bool) ([]*str
 		modelsCol = `"models"`
 	}
 
+	baseURLCol := "`base_url`"
+	// 如果是 PostgreSQL，使用双引号
+	if common.UsingPostgreSQL {
+		baseURLCol = `"base_url"`
+	}
+
 	order := "priority desc"
 	if idSort {
 		order = "id desc"
@@ -469,11 +491,11 @@ func SearchTags(keyword string, group string, model string, idSort bool) ([]*str
 			// sqlite, PostgreSQL
 			groupCondition = `(',' || ` + groupCol + ` || ',') LIKE ?`
 		}
-		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ?) AND " + modelsCol + ` LIKE ? AND ` + groupCondition
-		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+model+"%", "%,"+group+",%")
+		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ? OR " + baseURLCol + " LIKE ?) AND " + modelsCol + ` LIKE ? AND ` + groupCondition
+		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+keyword+"%", "%"+model+"%", "%,"+group+",%")
 	} else {
-		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ?) AND " + modelsCol + " LIKE ?"
-		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+model+"%")
+		whereClause = "(id = ? OR name LIKE ? OR " + keyCol + " = ? OR " + baseURLCol + " LIKE ?) AND " + modelsCol + " LIKE ?"
+		args = append(args, common.String2Int(keyword), "%"+keyword+"%", keyword, "%"+keyword+"%", "%"+model+"%")
 	}
 
 	subQuery := baseQuery.Where(whereClause, args...).

model/log.go

@@ -1,6 +1,7 @@
 package model
 
 import (
+	"context"
 	"fmt"
 	"one-api/common"
 	"os"
@@ -40,6 +41,7 @@ const (
 	LogTypeConsume
 	LogTypeManage
 	LogTypeSystem
+	LogTypeError
 )
 
 func formatUserLogs(logs []*Log) {
@@ -88,6 +90,35 @@ func RecordLog(userId int, logType int, content string) {
 	}
 }
 
+func RecordErrorLog(c *gin.Context, userId int, channelId int, modelName string, tokenName string, content string, tokenId int, useTimeSeconds int,
+	isStream bool, group string, other map[string]interface{}) {
+	common.LogInfo(c, fmt.Sprintf("record error log: userId=%d, channelId=%d, modelName=%s, tokenName=%s, content=%s", userId, channelId, modelName, tokenName, content))
+	username := c.GetString("username")
+	otherStr := common.MapToJsonStr(other)
+	log := &Log{
+		UserId:           userId,
+		Username:         username,
+		CreatedAt:        common.GetTimestamp(),
+		Type:             LogTypeError,
+		Content:          content,
+		PromptTokens:     0,
+		CompletionTokens: 0,
+		TokenName:        tokenName,
+		ModelName:        modelName,
+		Quota:            0,
+		ChannelId:        channelId,
+		TokenId:          tokenId,
+		UseTime:          useTimeSeconds,
+		IsStream:         isStream,
+		Group:            group,
+		Other:            otherStr,
+	}
+	err := LOG_DB.Create(log).Error
+	if err != nil {
+		common.LogError(c, "failed to record log: "+err.Error())
+	}
+}
+
 func RecordConsumeLog(c *gin.Context, userId int, channelId int, promptTokens int, completionTokens int,
 	modelName string, tokenName string, quota int, content string, tokenId int, userQuota int, useTimeSeconds int,
 	isStream bool, group string, other map[string]interface{}) {
@@ -310,7 +341,25 @@ func SumUsedToken(logType int, startTimestamp int64, endTimestamp int64, modelNa
 	return token
 }
 
-func DeleteOldLog(targetTimestamp int64) (int64, error) {
-	result := LOG_DB.Where("created_at < ?", targetTimestamp).Delete(&Log{})
-	return result.RowsAffected, result.Error
+func DeleteOldLog(ctx context.Context, targetTimestamp int64, limit int) (int64, error) {
+	var total int64 = 0
+
+	for {
+		if nil != ctx.Err() {
+			return total, ctx.Err()
+		}
+
+		result := LOG_DB.Where("created_at < ?", targetTimestamp).Limit(limit).Delete(&Log{})
+		if nil != result.Error {
+			return total, result.Error
+		}
+
+		total += result.RowsAffected
+
+		if result.RowsAffected < int64(limit) {
+			break
+		}
+	}
+
+	return total, nil
 }

model/option.go

@@ -67,6 +67,7 @@ func InitOptionMap() {
 	common.OptionMap["ServerAddress"] = ""
 	common.OptionMap["WorkerUrl"] = setting.WorkerUrl
 	common.OptionMap["WorkerValidKey"] = setting.WorkerValidKey
+	common.OptionMap["WorkerAllowHttpImageRequestEnabled"] = strconv.FormatBool(setting.WorkerAllowHttpImageRequestEnabled)
 	common.OptionMap["PayAddress"] = ""
 	common.OptionMap["CustomCallbackAddress"] = ""
 	common.OptionMap["EpayId"] = ""
@@ -92,6 +93,7 @@ func InitOptionMap() {
 	common.OptionMap["ModelRequestRateLimitCount"] = strconv.Itoa(setting.ModelRequestRateLimitCount)
 	common.OptionMap["ModelRequestRateLimitDurationMinutes"] = strconv.Itoa(setting.ModelRequestRateLimitDurationMinutes)
 	common.OptionMap["ModelRequestRateLimitSuccessCount"] = strconv.Itoa(setting.ModelRequestRateLimitSuccessCount)
+	common.OptionMap["ModelRequestRateLimitGroup"] = setting.ModelRequestRateLimitGroup2JSONString()
 	common.OptionMap["ModelRatio"] = operation_setting.ModelRatio2JSONString()
 	common.OptionMap["ModelPrice"] = operation_setting.ModelPrice2JSONString()
 	common.OptionMap["CacheRatio"] = operation_setting.CacheRatio2JSONString()
@@ -256,6 +258,8 @@ func updateOptionMap(key string, value string) (err error) {
 			setting.StopOnSensitiveEnabled = boolValue
 		case "SMTPSSLEnabled":
 			common.SMTPSSLEnabled = boolValue
+		case "WorkerAllowHttpImageRequestEnabled":
+			setting.WorkerAllowHttpImageRequestEnabled = boolValue
 		}
 	}
 	switch key {
@@ -338,6 +342,8 @@ func updateOptionMap(key string, value string) (err error) {
 		setting.ModelRequestRateLimitDurationMinutes, _ = strconv.Atoi(value)
 	case "ModelRequestRateLimitSuccessCount":
 		setting.ModelRequestRateLimitSuccessCount, _ = strconv.Atoi(value)
+	case "ModelRequestRateLimitGroup":
+		err = setting.UpdateModelRequestRateLimitGroupByJSONString(value)
 	case "RetryTimes":
 		common.RetryTimes, _ = strconv.Atoi(value)
 	case "DataExportInterval":

model/user.go

@@ -18,6 +18,7 @@ type User struct {
 	Id               int            `json:"id"`
 	Username         string         `json:"username" gorm:"unique;index" validate:"max=12"`
 	Password         string         `json:"password" gorm:"not null;" validate:"min=8,max=20"`
+	OriginalPassword string         `json:"original_password" gorm:"-:all"` // this field is only for Password change verification, don't save it to database!
 	DisplayName      string         `json:"display_name" gorm:"index" validate:"max=20"`
 	Role             int            `json:"role" gorm:"type:int;default:1"`   // admin, common
 	Status           int            `json:"status" gorm:"type:int;default:1"` // enabled, disabled
@@ -108,7 +109,7 @@ func CheckUserExistOrDeleted(username string, email string) (bool, error) {
 
 func GetMaxUserId() int {
 	var user User
-	DB.Last(&user)
+	DB.Unscoped().Last(&user)
 	return user.Id
 }
 

relay/channel/adapter.go

@@ -1,11 +1,12 @@
 package channel
 
 import (
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
 	relaycommon "one-api/relay/common"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor interface {
@@ -18,6 +19,7 @@ type Adaptor interface {
 	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)
 	ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.ImageRequest) (any, error)
+	ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error)
 	DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error)
 	DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode)
 	GetModelList() []string

relay/channel/ali/adaptor.go

@@ -3,7 +3,6 @@ package ali
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -11,6 +10,8 @@ import (
 	"one-api/relay/channel/openai"
 	relaycommon "one-api/relay/common"
 	"one-api/relay/constant"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -32,6 +33,8 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 		fullRequestURL = fmt.Sprintf("%s/api/v1/services/embeddings/text-embedding/text-embedding", info.BaseUrl)
 	case constant.RelayModeImagesGenerations:
 		fullRequestURL = fmt.Sprintf("%s/api/v1/services/aigc/text2image/image-synthesis", info.BaseUrl)
+	case constant.RelayModeCompletions:
+		fullRequestURL = fmt.Sprintf("%s/compatible-mode/v1/completions", info.BaseUrl)
 	default:
 		fullRequestURL = fmt.Sprintf("%s/compatible-mode/v1/chat/completions", info.BaseUrl)
 	}
@@ -79,6 +82,11 @@ func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInf
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/ali/constants.go

@@ -1,7 +1,12 @@
 package ali
 
 var ModelList = []string{
-	"qwen-turbo", "qwen-plus", "qwen-max", "qwen-max-longcontext",
+	"qwen-turbo",
+	"qwen-plus",
+	"qwen-max",
+	"qwen-max-longcontext",
+	"qwq-32b",
+	"qwen3-235b-a22b",
 	"text-embedding-v1",
 }
 

relay/channel/api_request.go

@@ -1,16 +1,23 @@
 package channel
 
 import (
+	"context"
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
-	"github.com/gorilla/websocket"
 	"io"
 	"net/http"
 	common2 "one-api/common"
 	"one-api/relay/common"
 	"one-api/relay/constant"
+	"one-api/relay/helper"
 	"one-api/service"
+	"one-api/setting/operation_setting"
+	"sync"
+	"time"
+
+	"github.com/bytedance/gopkg/util/gopool"
+	"github.com/gin-gonic/gin"
+	"github.com/gorilla/websocket"
 )
 
 func SetupApiRequestHeader(info *common.RelayInfo, c *gin.Context, req *http.Header) {
@@ -55,6 +62,9 @@ func DoFormRequest(a Adaptor, c *gin.Context, info *common.RelayInfo, requestBod
 	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)
@@ -105,7 +115,62 @@ func doRequest(c *gin.Context, req *http.Request, info *common.RelayInfo) (*http
 	} else {
 		client = service.GetHttpClient()
 	}
+	// 流式请求 ping 保活
+	var stopPinger func()
+	generalSettings := operation_setting.GetGeneralSetting()
+	pingEnabled := generalSettings.PingIntervalEnabled
+	var pingerWg sync.WaitGroup
+	if info.IsStream {
+		helper.SetEventStreamHeaders(c)
+		pingInterval := time.Duration(generalSettings.PingIntervalSeconds) * time.Second
+		var pingerCtx context.Context
+		pingerCtx, stopPinger = context.WithCancel(c.Request.Context())
+
+		if pingEnabled {
+			pingerWg.Add(1)
+			gopool.Go(func() {
+				defer pingerWg.Done()
+				if pingInterval <= 0 {
+					pingInterval = helper.DefaultPingInterval
+				}
+
+				ticker := time.NewTicker(pingInterval)
+				defer ticker.Stop()
+				var pingMutex sync.Mutex
+				if common2.DebugEnabled {
+					println("SSE ping goroutine started")
+				}
+
+				for {
+					select {
+					case <-ticker.C:
+						pingMutex.Lock()
+						err2 := helper.PingData(c)
+						pingMutex.Unlock()
+						if err2 != nil {
+							common2.LogError(c, "SSE ping error: "+err.Error())
+							return
+						}
+						if common2.DebugEnabled {
+							println("SSE ping data sent.")
+						}
+					case <-pingerCtx.Done():
+						if common2.DebugEnabled {
+							println("SSE ping goroutine stopped.")
+						}
+						return
+					}
+				}
+			})
+		}
+	}
+
 	resp, err := client.Do(req)
+	// request结束后停止ping
+	if info.IsStream && pingEnabled {
+		stopPinger()
+		pingerWg.Wait()
+	}
 	if err != nil {
 		return nil, err
 	}

relay/channel/aws/adaptor.go

@@ -2,13 +2,14 @@ package aws
 
 import (
 	"errors"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
 	"one-api/relay/channel/claude"
 	relaycommon "one-api/relay/common"
 	"one-api/setting/model_setting"
+
+	"github.com/gin-gonic/gin"
 )
 
 const (
@@ -74,6 +75,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	return nil, errors.New("not implemented")
+}
+
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
 	return nil, nil
 }

relay/channel/aws/constants.go

@@ -11,6 +11,8 @@ var awsModelIDMap = map[string]string{
 	"claude-3-5-sonnet-20241022": "anthropic.claude-3-5-sonnet-20241022-v2:0",
 	"claude-3-5-haiku-20241022":  "anthropic.claude-3-5-haiku-20241022-v1:0",
 	"claude-3-7-sonnet-20250219": "anthropic.claude-3-7-sonnet-20250219-v1:0",
+	"claude-sonnet-4-20250514":   "anthropic.claude-sonnet-4-20250514-v1:0",
+	"claude-opus-4-20250514":     "anthropic.claude-opus-4-20250514-v1:0",
 }
 
 var awsModelCanCrossRegionMap = map[string]map[string]bool{
@@ -41,6 +43,16 @@ var awsModelCanCrossRegionMap = map[string]map[string]bool{
 	},
 	"anthropic.claude-3-7-sonnet-20250219-v1:0": {
 		"us": true,
+		"ap": true,
+		"eu": true,
+	},
+	"anthropic.claude-sonnet-4-20250514-v1:0": {
+		"us": true,
+		"ap": true,
+		"eu": true,
+	},
+	"anthropic.claude-opus-4-20250514-v1:0": {
+		"us": true,
 	},
 }
 

relay/channel/baidu/adaptor.go

@@ -3,7 +3,6 @@ package baidu
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -11,6 +10,8 @@ import (
 	relaycommon "one-api/relay/common"
 	"one-api/relay/constant"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -130,6 +131,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return baiduEmbeddingRequest, nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/baidu_v2/adaptor.go

@@ -3,13 +3,14 @@ package baidu_v2
 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"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -60,6 +61,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/claude/adaptor.go

@@ -3,7 +3,6 @@ package claude
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -11,6 +10,8 @@ import (
 	relaycommon "one-api/relay/common"
 	"one-api/setting/model_setting"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 const (
@@ -37,10 +38,10 @@ func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInf
 }
 
 func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
-	if strings.HasPrefix(info.UpstreamModelName, "claude-3") {
-		a.RequestMode = RequestModeMessage
-	} else {
+	if strings.HasPrefix(info.UpstreamModelName, "claude-2") || strings.HasPrefix(info.UpstreamModelName, "claude-instant") {
 		a.RequestMode = RequestModeCompletion
+	} else {
+		a.RequestMode = RequestModeMessage
 	}
 }
 
@@ -84,6 +85,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/claude/constants.go

@@ -13,6 +13,10 @@ var ModelList = []string{
 	"claude-3-5-sonnet-20241022",
 	"claude-3-7-sonnet-20250219",
 	"claude-3-7-sonnet-20250219-thinking",
+	"claude-sonnet-4-20250514",
+	"claude-sonnet-4-20250514-thinking",
+	"claude-opus-4-20250514",
+	"claude-opus-4-20250514-thinking",
 }
 
 var ChannelName = "claude"

relay/channel/claude/relay-claude.go

@@ -300,6 +300,13 @@ func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *dto.ClaudeResponse
 	response.Model = claudeResponse.Model
 	response.Choices = make([]dto.ChatCompletionsStreamResponseChoice, 0)
 	tools := make([]dto.ToolCallResponse, 0)
+	fcIdx := 0
+	if claudeResponse.Index != nil {
+		fcIdx = *claudeResponse.Index - 1
+		if fcIdx < 0 {
+			fcIdx = 0
+		}
+	}
 	var choice dto.ChatCompletionsStreamResponseChoice
 	if reqMode == RequestModeCompletion {
 		choice.Delta.SetContentString(claudeResponse.Completion)
@@ -319,7 +326,7 @@ func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *dto.ClaudeResponse
 				//choice.Delta.SetContentString(claudeResponse.ContentBlock.Text)
 				if claudeResponse.ContentBlock.Type == "tool_use" {
 					tools = append(tools, dto.ToolCallResponse{
-						Index: common.GetPointer(0),
+						Index: common.GetPointer(fcIdx),
 						ID:    claudeResponse.ContentBlock.Id,
 						Type:  "function",
 						Function: dto.FunctionResponse{
@@ -338,7 +345,7 @@ func StreamResponseClaude2OpenAI(reqMode int, claudeResponse *dto.ClaudeResponse
 				case "input_json_delta":
 					tools = append(tools, dto.ToolCallResponse{
 						Type:  "function",
-						Index: common.GetPointer(0),
+						Index: common.GetPointer(fcIdx),
 						Function: dto.FunctionResponse{
 							Arguments: *claudeResponse.Delta.PartialJson,
 						},

relay/channel/cloudflare/adaptor.go

@@ -55,6 +55,11 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	}
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/cohere/adaptor.go

@@ -3,13 +3,14 @@ package cohere
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
 	"one-api/relay/channel"
 	relaycommon "one-api/relay/common"
 	"one-api/relay/constant"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -52,6 +53,11 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	return requestOpenAI2Cohere(*request), nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/coze/adaptor.go

@@ -0,0 +1,132 @@
+package coze
+
+import (
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"one-api/dto"
+	"one-api/relay/channel"
+	"one-api/relay/common"
+	"time"
+
+	"github.com/gin-gonic/gin"
+)
+
+type Adaptor struct {
+}
+
+// ConvertAudioRequest implements channel.Adaptor.
+func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *common.RelayInfo, request dto.AudioRequest) (io.Reader, error) {
+	return nil, errors.New("not implemented")
+}
+
+// ConvertClaudeRequest implements channel.Adaptor.
+func (a *Adaptor) ConvertClaudeRequest(c *gin.Context, info *common.RelayInfo, request *dto.ClaudeRequest) (any, error) {
+	return nil, errors.New("not implemented")
+}
+
+// ConvertEmbeddingRequest implements channel.Adaptor.
+func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *common.RelayInfo, request dto.EmbeddingRequest) (any, error) {
+	return nil, errors.New("not implemented")
+}
+
+// ConvertImageRequest implements channel.Adaptor.
+func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *common.RelayInfo, request dto.ImageRequest) (any, error) {
+	return nil, errors.New("not implemented")
+}
+
+// ConvertOpenAIRequest implements channel.Adaptor.
+func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *common.RelayInfo, request *dto.GeneralOpenAIRequest) (any, error) {
+	if request == nil {
+		return nil, errors.New("request is nil")
+	}
+	return convertCozeChatRequest(c, *request), nil
+}
+
+// ConvertOpenAIResponsesRequest implements channel.Adaptor.
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *common.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	return nil, errors.New("not implemented")
+}
+
+// ConvertRerankRequest implements channel.Adaptor.
+func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
+	return nil, errors.New("not implemented")
+}
+
+// DoRequest implements channel.Adaptor.
+func (a *Adaptor) DoRequest(c *gin.Context, info *common.RelayInfo, requestBody io.Reader) (any, error) {
+	if info.IsStream {
+		return channel.DoApiRequest(a, c, info, requestBody)
+	}
+	// 首先发送创建消息请求，成功后再发送获取消息请求
+	// 发送创建消息请求
+	resp, err := channel.DoApiRequest(a, c, info, requestBody)
+	if err != nil {
+		return nil, err
+	}
+	// 解析 resp
+	var cozeResponse CozeChatResponse
+	respBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return nil, err
+	}
+	err = json.Unmarshal(respBody, &cozeResponse)
+	if cozeResponse.Code != 0 {
+		return nil, errors.New(cozeResponse.Msg)
+	}
+	c.Set("coze_conversation_id", cozeResponse.Data.ConversationId)
+	c.Set("coze_chat_id", cozeResponse.Data.Id)
+	// 轮询检查消息是否完成
+	for {
+		err, isComplete := checkIfChatComplete(a, c, info)
+		if err != nil {
+			return nil, err
+		} else {
+			if isComplete {
+				break
+			}
+		}
+		time.Sleep(time.Second * 1)
+	}
+	// 发送获取消息请求
+	return getChatDetail(a, c, info)
+}
+
+// DoResponse implements channel.Adaptor.
+func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *common.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
+	if info.IsStream {
+		err, usage = cozeChatStreamHandler(c, resp, info)
+	} else {
+		err, usage = cozeChatHandler(c, resp, info)
+	}
+	return
+}
+
+// GetChannelName implements channel.Adaptor.
+func (a *Adaptor) GetChannelName() string {
+	return ChannelName
+}
+
+// GetModelList implements channel.Adaptor.
+func (a *Adaptor) GetModelList() []string {
+	return ModelList
+}
+
+// GetRequestURL implements channel.Adaptor.
+func (a *Adaptor) GetRequestURL(info *common.RelayInfo) (string, error) {
+	return fmt.Sprintf("%s/v3/chat", info.BaseUrl), nil
+}
+
+// Init implements channel.Adaptor.
+func (a *Adaptor) Init(info *common.RelayInfo) {
+
+}
+
+// SetupRequestHeader implements channel.Adaptor.
+func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *common.RelayInfo) error {
+	channel.SetupApiRequestHeader(info, c, req)
+	req.Set("Authorization", "Bearer "+info.ApiKey)
+	return nil
+}

relay/channel/coze/constants.go

@@ -0,0 +1,30 @@
+package coze
+
+var ModelList = []string{
+	"moonshot-v1-8k",
+	"moonshot-v1-32k",
+	"moonshot-v1-128k",
+	"Baichuan4",
+	"abab6.5s-chat-pro",
+	"glm-4-0520",
+	"qwen-max",
+	"deepseek-r1",
+	"deepseek-v3",
+	"deepseek-r1-distill-qwen-32b",
+	"deepseek-r1-distill-qwen-7b",
+	"step-1v-8k",
+	"step-1.5v-mini",
+	"Doubao-pro-32k",
+	"Doubao-pro-256k",
+	"Doubao-lite-128k",
+	"Doubao-lite-32k",
+	"Doubao-vision-lite-32k",
+	"Doubao-vision-pro-32k",
+	"Doubao-1.5-pro-vision-32k",
+	"Doubao-1.5-lite-32k",
+	"Doubao-1.5-pro-32k",
+	"Doubao-1.5-thinking-pro",
+	"Doubao-1.5-pro-256k",
+}
+
+var ChannelName = "coze"

relay/channel/coze/dto.go

@@ -0,0 +1,78 @@
+package coze
+
+import "encoding/json"
+
+type CozeError struct {
+	Code    int    `json:"code"`
+	Message string `json:"message"`
+}
+
+type CozeEnterMessage struct {
+	Role        string          `json:"role"`
+	Type        string          `json:"type,omitempty"`
+	Content     json.RawMessage `json:"content,omitempty"`
+	MetaData    json.RawMessage `json:"meta_data,omitempty"`
+	ContentType string          `json:"content_type,omitempty"`
+}
+
+type CozeChatRequest struct {
+	BotId              string             `json:"bot_id"`
+	UserId             string             `json:"user_id"`
+	AdditionalMessages []CozeEnterMessage `json:"additional_messages,omitempty"`
+	Stream             bool               `json:"stream,omitempty"`
+	CustomVariables    json.RawMessage    `json:"custom_variables,omitempty"`
+	AutoSaveHistory    bool               `json:"auto_save_history,omitempty"`
+	MetaData           json.RawMessage    `json:"meta_data,omitempty"`
+	ExtraParams        json.RawMessage    `json:"extra_params,omitempty"`
+	ShortcutCommand    json.RawMessage    `json:"shortcut_command,omitempty"`
+	Parameters         json.RawMessage    `json:"parameters,omitempty"`
+}
+
+type CozeChatResponse struct {
+	Code int                  `json:"code"`
+	Msg  string               `json:"msg"`
+	Data CozeChatResponseData `json:"data"`
+}
+
+type CozeChatResponseData struct {
+	Id             string        `json:"id"`
+	ConversationId string        `json:"conversation_id"`
+	BotId          string        `json:"bot_id"`
+	CreatedAt      int64         `json:"created_at"`
+	LastError      CozeError     `json:"last_error"`
+	Status         string        `json:"status"`
+	Usage          CozeChatUsage `json:"usage"`
+}
+
+type CozeChatUsage struct {
+	TokenCount  int `json:"token_count"`
+	OutputCount int `json:"output_count"`
+	InputCount  int `json:"input_count"`
+}
+
+type CozeChatDetailResponse struct {
+	Data   []CozeChatV3MessageDetail `json:"data"`
+	Code   int                       `json:"code"`
+	Msg    string                    `json:"msg"`
+	Detail CozeResponseDetail        `json:"detail"`
+}
+
+type CozeChatV3MessageDetail struct {
+	Id               string          `json:"id"`
+	Role             string          `json:"role"`
+	Type             string          `json:"type"`
+	BotId            string          `json:"bot_id"`
+	ChatId           string          `json:"chat_id"`
+	Content          json.RawMessage `json:"content"`
+	MetaData         json.RawMessage `json:"meta_data"`
+	CreatedAt        int64           `json:"created_at"`
+	SectionId        string          `json:"section_id"`
+	UpdatedAt        int64           `json:"updated_at"`
+	ContentType      string          `json:"content_type"`
+	ConversationId   string          `json:"conversation_id"`
+	ReasoningContent string          `json:"reasoning_content"`
+}
+
+type CozeResponseDetail struct {
+	Logid string `json:"logid"`
+}

relay/channel/coze/relay-coze.go

@@ -0,0 +1,300 @@
+package coze
+
+import (
+	"bufio"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"one-api/common"
+	"one-api/dto"
+	relaycommon "one-api/relay/common"
+	"one-api/relay/helper"
+	"one-api/service"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+)
+
+func convertCozeChatRequest(c *gin.Context, request dto.GeneralOpenAIRequest) *CozeChatRequest {
+	var messages []CozeEnterMessage
+	// 将 request的messages的role为user的content转换为CozeMessage
+	for _, message := range request.Messages {
+		if message.Role == "user" {
+			messages = append(messages, CozeEnterMessage{
+				Role:    "user",
+				Content: message.Content,
+				// TODO: support more content type
+				ContentType: "text",
+			})
+		}
+	}
+	user := request.User
+	if user == "" {
+		user = helper.GetResponseID(c)
+	}
+	cozeRequest := &CozeChatRequest{
+		BotId:              c.GetString("bot_id"),
+		UserId:             user,
+		AdditionalMessages: messages,
+		Stream:             request.Stream,
+	}
+	return cozeRequest
+}
+
+func cozeChatHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*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.OpenAIErrorWrapperLocal(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	// convert coze response to openai response
+	var response dto.TextResponse
+	var cozeResponse CozeChatDetailResponse
+	response.Model = info.UpstreamModelName
+	err = json.Unmarshal(responseBody, &cozeResponse)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if cozeResponse.Code != 0 {
+		return service.OpenAIErrorWrapper(errors.New(cozeResponse.Msg), fmt.Sprintf("%d", cozeResponse.Code), http.StatusInternalServerError), nil
+	}
+	// 从上下文获取 usage
+	var usage dto.Usage
+	usage.PromptTokens = c.GetInt("coze_input_count")
+	usage.CompletionTokens = c.GetInt("coze_output_count")
+	usage.TotalTokens = c.GetInt("coze_token_count")
+	response.Usage = usage
+	response.Id = helper.GetResponseID(c)
+
+	var responseContent json.RawMessage
+	for _, data := range cozeResponse.Data {
+		if data.Type == "answer" {
+			responseContent = data.Content
+			response.Created = data.CreatedAt
+		}
+	}
+	// 添加 response.Choices
+	response.Choices = []dto.OpenAITextResponseChoice{
+		{
+			Index:        0,
+			Message:      dto.Message{Role: "assistant", Content: responseContent},
+			FinishReason: "stop",
+		},
+	}
+	jsonResponse, err := json.Marshal(response)
+	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)
+	_, _ = c.Writer.Write(jsonResponse)
+
+	return nil, &usage
+}
+
+func cozeChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	scanner := bufio.NewScanner(resp.Body)
+	scanner.Split(bufio.ScanLines)
+	helper.SetEventStreamHeaders(c)
+	id := helper.GetResponseID(c)
+	var responseText string
+
+	var currentEvent string
+	var currentData string
+	var usage dto.Usage
+
+	for scanner.Scan() {
+		line := scanner.Text()
+
+		if line == "" {
+			if currentEvent != "" && currentData != "" {
+				// handle last event
+				handleCozeEvent(c, currentEvent, currentData, &responseText, &usage, id, info)
+				currentEvent = ""
+				currentData = ""
+			}
+			continue
+		}
+
+		if strings.HasPrefix(line, "event:") {
+			currentEvent = strings.TrimSpace(line[6:])
+			continue
+		}
+
+		if strings.HasPrefix(line, "data:") {
+			currentData = strings.TrimSpace(line[5:])
+			continue
+		}
+	}
+
+	// Last event
+	if currentEvent != "" && currentData != "" {
+		handleCozeEvent(c, currentEvent, currentData, &responseText, &usage, id, info)
+	}
+
+	if err := scanner.Err(); err != nil {
+		return service.OpenAIErrorWrapper(err, "stream_scanner_error", http.StatusInternalServerError), nil
+	}
+	helper.Done(c)
+
+	if usage.TotalTokens == 0 {
+		usage.PromptTokens = info.PromptTokens
+		usage.CompletionTokens, _ = service.CountTextToken("gpt-3.5-turbo", responseText)
+		usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
+	}
+
+	return nil, &usage
+}
+
+func handleCozeEvent(c *gin.Context, event string, data string, responseText *string, usage *dto.Usage, id string, info *relaycommon.RelayInfo) {
+	switch event {
+	case "conversation.chat.completed":
+		// 将 data 解析为 CozeChatResponseData
+		var chatData CozeChatResponseData
+		err := json.Unmarshal([]byte(data), &chatData)
+		if err != nil {
+			common.SysError("error_unmarshalling_stream_response: " + err.Error())
+			return
+		}
+
+		usage.PromptTokens = chatData.Usage.InputCount
+		usage.CompletionTokens = chatData.Usage.OutputCount
+		usage.TotalTokens = chatData.Usage.TokenCount
+
+		finishReason := "stop"
+		stopResponse := helper.GenerateStopResponse(id, common.GetTimestamp(), info.UpstreamModelName, finishReason)
+		helper.ObjectData(c, stopResponse)
+
+	case "conversation.message.delta":
+		// 将 data 解析为 CozeChatV3MessageDetail
+		var messageData CozeChatV3MessageDetail
+		err := json.Unmarshal([]byte(data), &messageData)
+		if err != nil {
+			common.SysError("error_unmarshalling_stream_response: " + err.Error())
+			return
+		}
+
+		var content string
+		err = json.Unmarshal(messageData.Content, &content)
+		if err != nil {
+			common.SysError("error_unmarshalling_stream_response: " + err.Error())
+			return
+		}
+
+		*responseText += content
+
+		openaiResponse := dto.ChatCompletionsStreamResponse{
+			Id:      id,
+			Object:  "chat.completion.chunk",
+			Created: common.GetTimestamp(),
+			Model:   info.UpstreamModelName,
+		}
+
+		choice := dto.ChatCompletionsStreamResponseChoice{
+			Index: 0,
+		}
+		choice.Delta.SetContentString(content)
+		openaiResponse.Choices = append(openaiResponse.Choices, choice)
+
+		helper.ObjectData(c, openaiResponse)
+
+	case "error":
+		var errorData CozeError
+		err := json.Unmarshal([]byte(data), &errorData)
+		if err != nil {
+			common.SysError("error_unmarshalling_stream_response: " + err.Error())
+			return
+		}
+
+		common.SysError(fmt.Sprintf("stream event error: ", errorData.Code, errorData.Message))
+	}
+}
+
+func checkIfChatComplete(a *Adaptor, c *gin.Context, info *relaycommon.RelayInfo) (error, bool) {
+	requestURL := fmt.Sprintf("%s/v3/chat/retrieve", info.BaseUrl)
+
+	requestURL = requestURL + "?conversation_id=" + c.GetString("coze_conversation_id") + "&chat_id=" + c.GetString("coze_chat_id")
+	// 将 conversationId和chatId作为参数发送get请求
+	req, err := http.NewRequest("GET", requestURL, nil)
+	if err != nil {
+		return err, false
+	}
+	err = a.SetupRequestHeader(c, &req.Header, info)
+	if err != nil {
+		return err, false
+	}
+
+	resp, err := doRequest(req, info) // 调用 doRequest
+	if err != nil {
+		return err, false
+	}
+	if resp == nil { // 确保在 doRequest 失败时 resp 不为 nil 导致 panic
+		return fmt.Errorf("resp is nil"), false
+	}
+	defer resp.Body.Close() // 确保响应体被关闭
+
+	// 解析 resp 到 CozeChatResponse
+	var cozeResponse CozeChatResponse
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return fmt.Errorf("read response body failed: %w", err), false
+	}
+	err = json.Unmarshal(responseBody, &cozeResponse)
+	if err != nil {
+		return fmt.Errorf("unmarshal response body failed: %w", err), false
+	}
+	if cozeResponse.Data.Status == "completed" {
+		// 在上下文设置 usage
+		c.Set("coze_token_count", cozeResponse.Data.Usage.TokenCount)
+		c.Set("coze_output_count", cozeResponse.Data.Usage.OutputCount)
+		c.Set("coze_input_count", cozeResponse.Data.Usage.InputCount)
+		return nil, true
+	} else if cozeResponse.Data.Status == "failed" || cozeResponse.Data.Status == "canceled" || cozeResponse.Data.Status == "requires_action" {
+		return fmt.Errorf("chat status: %s", cozeResponse.Data.Status), false
+	} else {
+		return nil, false
+	}
+}
+
+func getChatDetail(a *Adaptor, c *gin.Context, info *relaycommon.RelayInfo) (*http.Response, error) {
+	requestURL := fmt.Sprintf("%s/v3/chat/message/list", info.BaseUrl)
+
+	requestURL = requestURL + "?conversation_id=" + c.GetString("coze_conversation_id") + "&chat_id=" + c.GetString("coze_chat_id")
+	req, err := http.NewRequest("GET", requestURL, nil)
+	if err != nil {
+		return nil, fmt.Errorf("new request failed: %w", err)
+	}
+	err = a.SetupRequestHeader(c, &req.Header, info)
+	if err != nil {
+		return nil, fmt.Errorf("setup request header failed: %w", err)
+	}
+	resp, err := doRequest(req, info)
+	if err != nil {
+		return nil, fmt.Errorf("do request failed: %w", err)
+	}
+	return resp, nil
+}
+
+func doRequest(req *http.Request, info *relaycommon.RelayInfo) (*http.Response, error) {
+	var client *http.Client
+	var err error // 声明 err 变量
+	if proxyURL, ok := info.ChannelSetting["proxy"]; ok {
+		client, err = service.NewProxyHttpClient(proxyURL.(string))
+		if err != nil {
+			return nil, fmt.Errorf("new proxy http client failed: %w", err)
+		}
+	} else {
+		client = service.GetHttpClient()
+	}
+	resp, err := client.Do(req)
+	if err != nil { // 增加对 client.Do(req) 返回错误的检查
+		return nil, fmt.Errorf("client.Do failed: %w", err)
+	}
+	// _ = resp.Body.Close()
+	return resp, nil
+}

relay/channel/deepseek/adaptor.go

@@ -3,7 +3,6 @@ package deepseek
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -11,6 +10,9 @@ import (
 	"one-api/relay/channel/openai"
 	relaycommon "one-api/relay/common"
 	"one-api/relay/constant"
+	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -36,9 +38,13 @@ func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
 }
 
 func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
+	fimBaseUrl := info.BaseUrl
+	if !strings.HasSuffix(info.BaseUrl, "/beta") {
+		fimBaseUrl += "/beta"
+	}
 	switch info.RelayMode {
 	case constant.RelayModeCompletions:
-		return fmt.Sprintf("%s/beta/completions", info.BaseUrl), nil
+		return fmt.Sprintf("%s/completions", fimBaseUrl), nil
 	default:
 		return fmt.Sprintf("%s/v1/chat/completions", info.BaseUrl), nil
 	}
@@ -66,6 +72,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/dify/adaptor.go

@@ -3,12 +3,13 @@ package dify
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
 	"one-api/relay/channel"
 	relaycommon "one-api/relay/common"
+
+	"github.com/gin-gonic/gin"
 )
 
 const (
@@ -86,6 +87,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/gemini/adaptor.go

@@ -12,7 +12,6 @@ import (
 	relaycommon "one-api/relay/common"
 	"one-api/service"
 	"one-api/setting/model_setting"
-
 	"strings"
 
 	"github.com/gin-gonic/gin"
@@ -70,6 +69,16 @@ func (a *Adaptor) Init(info *relaycommon.RelayInfo) {
 }
 
 func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
+
+	if model_setting.GetGeminiSettings().ThinkingAdapterEnabled {
+		// suffix -thinking and -nothinking
+		if strings.HasSuffix(info.OriginModelName, "-thinking") {
+			info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-thinking")
+		} else if strings.HasSuffix(info.OriginModelName, "-nothinking") {
+			info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-nothinking")
+		}
+	}
+
 	version := model_setting.GetGeminiVersionSetting(info.UpstreamModelName)
 
 	if strings.HasPrefix(info.UpstreamModelName, "imagen") {
@@ -99,11 +108,13 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	if request == nil {
 		return nil, errors.New("request is nil")
 	}
-	ai, err := CovertGemini2OpenAI(*request)
+
+	geminiRequest, err := CovertGemini2OpenAI(*request, info)
 	if err != nil {
 		return nil, err
 	}
-	return ai, nil
+
+	return geminiRequest, nil
 }
 
 func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dto.RerankRequest) (any, error) {
@@ -144,6 +155,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return geminiRequest, nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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)
 }
@@ -165,6 +181,18 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 	} else {
 		err, usage = GeminiChatHandler(c, resp, info)
 	}
+
+	//if usage.(*dto.Usage).CompletionTokenDetails.ReasoningTokens > 100 {
+	//	// 没有请求-thinking的情况下，产生思考token，则按照思考模型计费
+	//	if !strings.HasSuffix(info.OriginModelName, "-thinking") &&
+	//		!strings.HasSuffix(info.OriginModelName, "-nothinking") {
+	//		thinkingModelName := info.OriginModelName + "-thinking"
+	//		if operation_setting.SelfUseModeEnabled || helper.ContainPriceOrRatio(thinkingModelName) {
+	//			info.OriginModelName = thinkingModelName
+	//		}
+	//	}
+	//}
+
 	return
 }
 

relay/channel/gemini/dto.go

@@ -2,10 +2,19 @@ package gemini
 
 type GeminiChatRequest struct {
 	Contents           []GeminiChatContent        `json:"contents"`
-	SafetySettings     []GeminiChatSafetySettings `json:"safety_settings,omitempty"`
-	GenerationConfig   GeminiChatGenerationConfig `json:"generation_config,omitempty"`
+	SafetySettings     []GeminiChatSafetySettings `json:"safetySettings,omitempty"`
+	GenerationConfig   GeminiChatGenerationConfig `json:"generationConfig,omitempty"`
 	Tools              []GeminiChatTool           `json:"tools,omitempty"`
-	SystemInstructions *GeminiChatContent         `json:"system_instruction,omitempty"`
+	SystemInstructions *GeminiChatContent         `json:"systemInstruction,omitempty"`
+}
+
+type GeminiThinkingConfig struct {
+	IncludeThoughts bool `json:"includeThoughts,omitempty"`
+	ThinkingBudget  *int `json:"thinkingBudget,omitempty"`
+}
+
+func (c *GeminiThinkingConfig) SetThinkingBudget(budget int) {
+	c.ThinkingBudget = &budget
 }
 
 type GeminiInlineData struct {
@@ -45,6 +54,7 @@ type GeminiFileData struct {
 
 type GeminiPart struct {
 	Text                string                         `json:"text,omitempty"`
+	Thought             bool                           `json:"thought,omitempty"`
 	InlineData          *GeminiInlineData              `json:"inlineData,omitempty"`
 	FunctionCall        *FunctionCall                  `json:"functionCall,omitempty"`
 	FunctionResponse    *FunctionResponse              `json:"functionResponse,omitempty"`
@@ -71,15 +81,17 @@ type GeminiChatTool struct {
 }
 
 type GeminiChatGenerationConfig struct {
-	Temperature      *float64 `json:"temperature,omitempty"`
-	TopP             float64  `json:"topP,omitempty"`
-	TopK             float64  `json:"topK,omitempty"`
-	MaxOutputTokens  uint     `json:"maxOutputTokens,omitempty"`
-	CandidateCount   int      `json:"candidateCount,omitempty"`
-	StopSequences    []string `json:"stopSequences,omitempty"`
-	ResponseMimeType string   `json:"responseMimeType,omitempty"`
-	ResponseSchema   any      `json:"responseSchema,omitempty"`
-	Seed             int64    `json:"seed,omitempty"`
+	Temperature        *float64              `json:"temperature,omitempty"`
+	TopP               float64               `json:"topP,omitempty"`
+	TopK               float64               `json:"topK,omitempty"`
+	MaxOutputTokens    uint                  `json:"maxOutputTokens,omitempty"`
+	CandidateCount     int                   `json:"candidateCount,omitempty"`
+	StopSequences      []string              `json:"stopSequences,omitempty"`
+	ResponseMimeType   string                `json:"responseMimeType,omitempty"`
+	ResponseSchema     any                   `json:"responseSchema,omitempty"`
+	Seed               int64                 `json:"seed,omitempty"`
+	ResponseModalities []string              `json:"responseModalities,omitempty"`
+	ThinkingConfig     *GeminiThinkingConfig `json:"thinkingConfig,omitempty"`
 }
 
 type GeminiChatCandidate struct {
@@ -108,6 +120,7 @@ type GeminiUsageMetadata struct {
 	PromptTokenCount     int `json:"promptTokenCount"`
 	CandidatesTokenCount int `json:"candidatesTokenCount"`
 	TotalTokenCount      int `json:"totalTokenCount"`
+	ThoughtsTokenCount   int `json:"thoughtsTokenCount"`
 }
 
 // Imagen related structs

relay/channel/gemini/relay-gemini.go

@@ -19,11 +19,10 @@ import (
 )
 
 // Setting safety to the lowest possible values since Gemini is already powerless enough
-func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) (*GeminiChatRequest, error) {
+func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest, info *relaycommon.RelayInfo) (*GeminiChatRequest, error) {
 
 	geminiRequest := GeminiChatRequest{
 		Contents: make([]GeminiChatContent, 0, len(textRequest.Messages)),
-		//SafetySettings: []GeminiChatSafetySettings{},
 		GenerationConfig: GeminiChatGenerationConfig{
 			Temperature:     textRequest.Temperature,
 			TopP:            textRequest.TopP,
@@ -32,6 +31,30 @@ func CovertGemini2OpenAI(textRequest dto.GeneralOpenAIRequest) (*GeminiChatReque
 		},
 	}
 
+	if model_setting.IsGeminiModelSupportImagine(info.UpstreamModelName) {
+		geminiRequest.GenerationConfig.ResponseModalities = []string{
+			"TEXT",
+			"IMAGE",
+		}
+	}
+
+	if model_setting.GetGeminiSettings().ThinkingAdapterEnabled {
+		if strings.HasSuffix(info.OriginModelName, "-thinking") {
+			budgetTokens := model_setting.GetGeminiSettings().ThinkingAdapterBudgetTokensPercentage * float64(geminiRequest.GenerationConfig.MaxOutputTokens)
+			if budgetTokens == 0 || budgetTokens > 24576 {
+				budgetTokens = 24576
+			}
+			geminiRequest.GenerationConfig.ThinkingConfig = &GeminiThinkingConfig{
+				ThinkingBudget:  common.GetPointer(int(budgetTokens)),
+				IncludeThoughts: true,
+			}
+		} else if strings.HasSuffix(info.OriginModelName, "-nothinking") {
+			geminiRequest.GenerationConfig.ThinkingConfig = &GeminiThinkingConfig{
+				ThinkingBudget: common.GetPointer(0),
+			}
+		}
+	}
+
 	safetySettings := make([]GeminiChatSafetySettings, 0, len(SafetySettingList))
 	for _, category := range SafetySettingList {
 		safetySettings = append(safetySettings, GeminiChatSafetySettings{
@@ -279,6 +302,13 @@ func cleanFunctionParameters(params interface{}) interface{} {
 		cleanedMap[k] = v
 	}
 
+	// Remove unsupported root-level fields
+	delete(cleanedMap, "default")
+	delete(cleanedMap, "exclusiveMaximum")
+	delete(cleanedMap, "exclusiveMinimum")
+	delete(cleanedMap, "$schema")
+	delete(cleanedMap, "additionalProperties")
+
 	// Clean properties
 	if props, ok := cleanedMap["properties"].(map[string]interface{}); ok && props != nil {
 		cleanedProps := make(map[string]interface{})
@@ -299,6 +329,8 @@ func cleanFunctionParameters(params interface{}) interface{} {
 			delete(cleanedPropMap, "default")
 			delete(cleanedPropMap, "exclusiveMaximum")
 			delete(cleanedPropMap, "exclusiveMinimum")
+			delete(cleanedPropMap, "$schema")
+			delete(cleanedPropMap, "additionalProperties")
 
 			// Check and clean 'format' for string types
 			if propType, typeExists := cleanedPropMap["type"].(string); typeExists && propType == "string" {
@@ -359,6 +391,7 @@ func removeAdditionalPropertiesWithDepth(schema interface{}, depth int) interfac
 	}
 	// 删除所有的title字段
 	delete(v, "title")
+	delete(v, "$schema")
 	// 如果type不为object和array，则直接返回
 	if typeVal, exists := v["type"]; !exists || (typeVal != "object" && typeVal != "array") {
 		return schema
@@ -506,6 +539,8 @@ func responseGeminiChat2OpenAI(response *GeminiChatResponse) *dto.OpenAITextResp
 					if call := getResponseToolCall(&part); call != nil {
 						toolCalls = append(toolCalls, *call)
 					}
+				} else if part.Thought {
+					choice.Message.ReasoningContent = part.Text
 				} else {
 					if part.ExecutableCode != nil {
 						texts = append(texts, "```"+part.ExecutableCode.Language+"\n"+part.ExecutableCode.Code+"\n```")
@@ -523,7 +558,6 @@ func responseGeminiChat2OpenAI(response *GeminiChatResponse) *dto.OpenAITextResp
 				choice.Message.SetToolCalls(toolCalls)
 				isToolCall = true
 			}
-
 			choice.Message.SetStringContent(strings.Join(texts, "\n"))
 
 		}
@@ -546,9 +580,10 @@ func responseGeminiChat2OpenAI(response *GeminiChatResponse) *dto.OpenAITextResp
 	return &fullTextResponse
 }
 
-func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) (*dto.ChatCompletionsStreamResponse, bool) {
+func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) (*dto.ChatCompletionsStreamResponse, bool, bool) {
 	choices := make([]dto.ChatCompletionsStreamResponseChoice, 0, len(geminiResponse.Candidates))
 	isStop := false
+	hasImage := false
 	for _, candidate := range geminiResponse.Candidates {
 		if candidate.FinishReason != nil && *candidate.FinishReason == "STOP" {
 			isStop = true
@@ -562,6 +597,7 @@ func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) (*dto.C
 		}
 		var texts []string
 		isTools := false
+		isThought := false
 		if candidate.FinishReason != nil {
 			// p := GeminiConvertFinishReason(*candidate.FinishReason)
 			switch *candidate.FinishReason {
@@ -574,12 +610,21 @@ func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) (*dto.C
 			}
 		}
 		for _, part := range candidate.Content.Parts {
-			if part.FunctionCall != nil {
+			if part.InlineData != nil {
+				if strings.HasPrefix(part.InlineData.MimeType, "image") {
+					imgText := "![image](data:" + part.InlineData.MimeType + ";base64," + part.InlineData.Data + ")"
+					texts = append(texts, imgText)
+					hasImage = true
+				}
+			} else if part.FunctionCall != nil {
 				isTools = true
 				if call := getResponseToolCall(&part); call != nil {
 					call.SetIndex(len(choice.Delta.ToolCalls))
 					choice.Delta.ToolCalls = append(choice.Delta.ToolCalls, *call)
 				}
+			} else if part.Thought {
+				isThought = true
+				texts = append(texts, part.Text)
 			} else {
 				if part.ExecutableCode != nil {
 					texts = append(texts, "```"+part.ExecutableCode.Language+"\n"+part.ExecutableCode.Code+"\n```\n")
@@ -592,7 +637,11 @@ func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) (*dto.C
 				}
 			}
 		}
-		choice.Delta.SetContentString(strings.Join(texts, "\n"))
+		if isThought {
+			choice.Delta.SetReasoningContent(strings.Join(texts, "\n"))
+		} else {
+			choice.Delta.SetContentString(strings.Join(texts, "\n"))
+		}
 		if isTools {
 			choice.FinishReason = &constant.FinishReasonToolCalls
 		}
@@ -602,7 +651,7 @@ func streamResponseGeminiChat2OpenAI(geminiResponse *GeminiChatResponse) (*dto.C
 	var response dto.ChatCompletionsStreamResponse
 	response.Object = "chat.completion.chunk"
 	response.Choices = choices
-	return &response, isStop
+	return &response, isStop, hasImage
 }
 
 func GeminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
@@ -610,23 +659,28 @@ func GeminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycom
 	id := fmt.Sprintf("chatcmpl-%s", common.GetUUID())
 	createAt := common.GetTimestamp()
 	var usage = &dto.Usage{}
+	var imageCount int
 
 	helper.StreamScannerHandler(c, resp, info, func(data string) bool {
 		var geminiResponse GeminiChatResponse
-		err := json.Unmarshal([]byte(data), &geminiResponse)
+		err := common.DecodeJsonStr(data, &geminiResponse)
 		if err != nil {
 			common.LogError(c, "error unmarshalling stream response: "+err.Error())
 			return false
 		}
 
-		response, isStop := streamResponseGeminiChat2OpenAI(&geminiResponse)
+		response, isStop, hasImage := streamResponseGeminiChat2OpenAI(&geminiResponse)
+		if hasImage {
+			imageCount++
+		}
 		response.Id = id
 		response.Created = createAt
 		response.Model = info.UpstreamModelName
-		// responseText += response.Choices[0].Delta.GetContentString()
 		if geminiResponse.UsageMetadata.TotalTokenCount != 0 {
 			usage.PromptTokens = geminiResponse.UsageMetadata.PromptTokenCount
 			usage.CompletionTokens = geminiResponse.UsageMetadata.CandidatesTokenCount
+			usage.CompletionTokenDetails.ReasoningTokens = geminiResponse.UsageMetadata.ThoughtsTokenCount
+			usage.TotalTokens = geminiResponse.UsageMetadata.TotalTokenCount
 		}
 		err = helper.ObjectData(c, response)
 		if err != nil {
@@ -641,9 +695,14 @@ func GeminiChatStreamHandler(c *gin.Context, resp *http.Response, info *relaycom
 
 	var response *dto.ChatCompletionsStreamResponse
 
-	usage.TotalTokens = usage.PromptTokens + usage.CompletionTokens
+	if imageCount != 0 {
+		if usage.CompletionTokens == 0 {
+			usage.CompletionTokens = imageCount * 258
+		}
+	}
+
 	usage.PromptTokensDetails.TextTokens = usage.PromptTokens
-	usage.CompletionTokenDetails.TextTokens = usage.CompletionTokens
+	usage.CompletionTokens = usage.TotalTokens - usage.PromptTokens
 
 	if info.ShouldIncludeUsage {
 		response = helper.GenerateFinalUsageResponse(id, createAt, info.UpstreamModelName, *usage)
@@ -666,8 +725,11 @@ func GeminiChatHandler(c *gin.Context, resp *http.Response, info *relaycommon.Re
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
 	}
+	if common.DebugEnabled {
+		println(string(responseBody))
+	}
 	var geminiResponse GeminiChatResponse
-	err = json.Unmarshal(responseBody, &geminiResponse)
+	err = common.DecodeJson(responseBody, &geminiResponse)
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
 	}
@@ -689,6 +751,10 @@ func GeminiChatHandler(c *gin.Context, resp *http.Response, info *relaycommon.Re
 		CompletionTokens: geminiResponse.UsageMetadata.CandidatesTokenCount,
 		TotalTokens:      geminiResponse.UsageMetadata.TotalTokenCount,
 	}
+
+	usage.CompletionTokenDetails.ReasoningTokens = geminiResponse.UsageMetadata.ThoughtsTokenCount
+	usage.CompletionTokens = usage.TotalTokens - usage.PromptTokens
+
 	fullTextResponse.Usage = usage
 	jsonResponse, err := json.Marshal(fullTextResponse)
 	if err != nil {

relay/channel/jina/adaptor.go

@@ -3,7 +3,6 @@ package jina
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -12,6 +11,8 @@ import (
 	relaycommon "one-api/relay/common"
 	"one-api/relay/common_handler"
 	"one-api/relay/constant"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -55,6 +56,11 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	return request, nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/mistral/adaptor.go

@@ -2,13 +2,14 @@ package mistral
 
 import (
 	"errors"
-	"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"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -59,6 +60,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/mokaai/adaptor.go

@@ -3,7 +3,6 @@ package mokaai
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -11,6 +10,8 @@ import (
 	relaycommon "one-api/relay/common"
 	"one-api/relay/constant"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -74,6 +75,11 @@ func (a *Adaptor) ConvertRerankRequest(c *gin.Context, relayMode int, request dt
 	return nil, nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/ollama/adaptor.go

@@ -2,7 +2,6 @@ package ollama
 
 import (
 	"errors"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -10,6 +9,8 @@ import (
 	"one-api/relay/channel/openai"
 	relaycommon "one-api/relay/common"
 	relayconstant "one-api/relay/constant"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -64,6 +65,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return requestOpenAI2Embeddings(request), nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/openai/adaptor.go

@@ -8,6 +8,7 @@ import (
 	"io"
 	"mime/multipart"
 	"net/http"
+	"net/textproto"
 	"one-api/common"
 	constant2 "one-api/constant"
 	"one-api/dto"
@@ -22,6 +23,7 @@ import (
 	"one-api/relay/common_handler"
 	"one-api/relay/constant"
 	"one-api/service"
+	"path/filepath"
 	"strings"
 
 	"github.com/gin-gonic/gin"
@@ -87,7 +89,10 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 		requestURL = fmt.Sprintf("%s?api-version=%s", requestURL, apiVersion)
 		task := strings.TrimPrefix(requestURL, "/v1/")
 		model_ := info.UpstreamModelName
-		model_ = strings.Replace(model_, ".", "", -1)
+		// 2025年5月10日后创建的渠道不移除.
+		if info.ChannelCreateTime < constant2.AzureNoRemoveDotTime {
+			model_ = strings.Replace(model_, ".", "", -1)
+		}
 		// https://github.com/songquanpeng/one-api/issues/67
 		requestURL = fmt.Sprintf("/openai/deployments/%s/%s", model_, task)
 		if info.RelayMode == constant.RelayModeRealtime {
@@ -147,14 +152,12 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	if info.ChannelType != common.ChannelTypeOpenAI && info.ChannelType != common.ChannelTypeAzure {
 		request.StreamOptions = nil
 	}
-	if strings.HasPrefix(request.Model, "o1") || strings.HasPrefix(request.Model, "o3") {
+	if strings.HasPrefix(request.Model, "o") {
 		if request.MaxCompletionTokens == 0 && request.MaxTokens != 0 {
 			request.MaxCompletionTokens = request.MaxTokens
 			request.MaxTokens = 0
 		}
-		if strings.HasPrefix(request.Model, "o3") || strings.HasPrefix(request.Model, "o1") {
-			request.Temperature = nil
-		}
+		request.Temperature = nil
 		if strings.HasSuffix(request.Model, "-high") {
 			request.ReasoningEffort = "high"
 			request.Model = strings.TrimSuffix(request.Model, "-high")
@@ -167,11 +170,13 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 		}
 		info.ReasoningEffort = request.ReasoningEffort
 		info.UpstreamModelName = request.Model
-	}
-	if request.Model == "o1" || request.Model == "o1-2024-12-17" || strings.HasPrefix(request.Model, "o3") {
-		//修改第一个Message的内容，将system改为developer
-		if len(request.Messages) > 0 && request.Messages[0].Role == "system" {
-			request.Messages[0].Role = "developer"
+
+		// o系列模型developer适配（o1-mini除外）
+		if !strings.HasPrefix(request.Model, "o1-mini") && !strings.HasPrefix(request.Model, "o1-preview") {
+			//修改第一个Message的内容，将system改为developer
+			if len(request.Messages) > 0 && request.Messages[0].Role == "system" {
+				request.Messages[0].Role = "developer"
+			}
 		}
 	}
 
@@ -236,11 +241,167 @@ func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInf
 }
 
 func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.ImageRequest) (any, error) {
+	switch info.RelayMode {
+	case constant.RelayModeImagesEdits:
+
+		var requestBody bytes.Buffer
+		writer := multipart.NewWriter(&requestBody)
+
+		writer.WriteField("model", request.Model)
+		// 获取所有表单字段
+		formData := c.Request.PostForm
+		// 遍历表单字段并打印输出
+		for key, values := range formData {
+			if key == "model" {
+				continue
+			}
+			for _, value := range values {
+				writer.WriteField(key, value)
+			}
+		}
+
+		// Parse the multipart form to handle both single image and multiple images
+		if err := c.Request.ParseMultipartForm(32 << 20); err != nil { // 32MB max memory
+			return nil, errors.New("failed to parse multipart form")
+		}
+
+		if c.Request.MultipartForm != nil && c.Request.MultipartForm.File != nil {
+			// Check if "image" field exists in any form, including array notation
+			var imageFiles []*multipart.FileHeader
+			var exists bool
+
+			// First check for standard "image" field
+			if imageFiles, exists = c.Request.MultipartForm.File["image"]; !exists || len(imageFiles) == 0 {
+				// If not found, check for "image[]" field
+				if imageFiles, exists = c.Request.MultipartForm.File["image[]"]; !exists || len(imageFiles) == 0 {
+					// If still not found, iterate through all fields to find any that start with "image["
+					foundArrayImages := false
+					for fieldName, files := range c.Request.MultipartForm.File {
+						if strings.HasPrefix(fieldName, "image[") && len(files) > 0 {
+							foundArrayImages = true
+							for _, file := range files {
+								imageFiles = append(imageFiles, file)
+							}
+						}
+					}
+
+					// If no image fields found at all
+					if !foundArrayImages && (len(imageFiles) == 0) {
+						return nil, errors.New("image is required")
+					}
+				}
+			}
+
+			// Process all image files
+			for i, fileHeader := range imageFiles {
+				file, err := fileHeader.Open()
+				if err != nil {
+					return nil, fmt.Errorf("failed to open image file %d: %w", i, err)
+				}
+				defer file.Close()
+
+				// If multiple images, use image[] as the field name
+				fieldName := "image"
+				if len(imageFiles) > 1 {
+					fieldName = "image[]"
+				}
+
+				// Determine MIME type based on file extension
+				mimeType := detectImageMimeType(fileHeader.Filename)
+
+				// Create a form file with the appropriate content type
+				h := make(textproto.MIMEHeader)
+				h.Set("Content-Disposition", fmt.Sprintf(`form-data; name="%s"; filename="%s"`, fieldName, fileHeader.Filename))
+				h.Set("Content-Type", mimeType)
+
+				part, err := writer.CreatePart(h)
+				if err != nil {
+					return nil, fmt.Errorf("create form part failed for image %d: %w", i, err)
+				}
+
+				if _, err := io.Copy(part, file); err != nil {
+					return nil, fmt.Errorf("copy file failed for image %d: %w", i, err)
+				}
+			}
+
+			// Handle mask file if present
+			if maskFiles, exists := c.Request.MultipartForm.File["mask"]; exists && len(maskFiles) > 0 {
+				maskFile, err := maskFiles[0].Open()
+				if err != nil {
+					return nil, errors.New("failed to open mask file")
+				}
+				defer maskFile.Close()
+
+				// Determine MIME type for mask file
+				mimeType := detectImageMimeType(maskFiles[0].Filename)
+
+				// Create a form file with the appropriate content type
+				h := make(textproto.MIMEHeader)
+				h.Set("Content-Disposition", fmt.Sprintf(`form-data; name="mask"; filename="%s"`, maskFiles[0].Filename))
+				h.Set("Content-Type", mimeType)
+
+				maskPart, err := writer.CreatePart(h)
+				if err != nil {
+					return nil, errors.New("create form file failed for mask")
+				}
+
+				if _, err := io.Copy(maskPart, maskFile); err != nil {
+					return nil, errors.New("copy mask file failed")
+				}
+			}
+		} else {
+			return nil, errors.New("no multipart form data found")
+		}
+
+		// 关闭 multipart 编写器以设置分界线
+		writer.Close()
+		c.Request.Header.Set("Content-Type", writer.FormDataContentType())
+		return bytes.NewReader(requestBody.Bytes()), nil
+
+	default:
+		return request, nil
+	}
+}
+
+// detectImageMimeType determines the MIME type based on the file extension
+func detectImageMimeType(filename string) string {
+	ext := strings.ToLower(filepath.Ext(filename))
+	switch ext {
+	case ".jpg", ".jpeg":
+		return "image/jpeg"
+	case ".png":
+		return "image/png"
+	case ".webp":
+		return "image/webp"
+	default:
+		// Try to detect from extension if possible
+		if strings.HasPrefix(ext, ".jp") {
+			return "image/jpeg"
+		}
+		// Default to png as a fallback
+		return "image/png"
+	}
+}
+
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// 模型后缀转换 reasoning effort
+	if strings.HasSuffix(request.Model, "-high") {
+		request.Reasoning.Effort = "high"
+		request.Model = strings.TrimSuffix(request.Model, "-high")
+	} else if strings.HasSuffix(request.Model, "-low") {
+		request.Reasoning.Effort = "low"
+		request.Model = strings.TrimSuffix(request.Model, "-low")
+	} else if strings.HasSuffix(request.Model, "-medium") {
+		request.Reasoning.Effort = "medium"
+		request.Model = strings.TrimSuffix(request.Model, "-medium")
+	}
 	return request, nil
 }
 
 func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, requestBody io.Reader) (any, error) {
-	if info.RelayMode == constant.RelayModeAudioTranscription || info.RelayMode == constant.RelayModeAudioTranslation {
+	if info.RelayMode == constant.RelayModeAudioTranscription ||
+		info.RelayMode == constant.RelayModeAudioTranslation ||
+		info.RelayMode == constant.RelayModeImagesEdits {
 		return channel.DoFormRequest(a, c, info, requestBody)
 	} else if info.RelayMode == constant.RelayModeRealtime {
 		return channel.DoWssRequest(a, c, info, requestBody)
@@ -259,10 +420,16 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 		fallthrough
 	case constant.RelayModeAudioTranscription:
 		err, usage = OpenaiSTTHandler(c, resp, info, a.ResponseFormat)
-	case constant.RelayModeImagesGenerations:
-		err, usage = OpenaiTTSHandler(c, resp, info)
+	case constant.RelayModeImagesGenerations, constant.RelayModeImagesEdits:
+		err, usage = OpenaiHandlerWithUsage(c, resp, info)
 	case constant.RelayModeRerank:
 		err, usage = common_handler.RerankHandler(c, info, resp)
+	case constant.RelayModeResponses:
+		if info.IsStream {
+			err, usage = OaiResponsesStreamHandler(c, resp, info)
+		} else {
+			err, usage = OaiResponsesHandler(c, resp, info)
+		}
 	default:
 		if info.IsStream {
 			err, usage = OaiStreamHandler(c, resp, info)

relay/channel/openai/helper.go

@@ -187,3 +187,10 @@ func handleFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, lastStream
 		}
 	}
 }
+
+func sendResponsesStreamData(c *gin.Context, streamResponse dto.ResponsesStreamResponse, data string) {
+	if data == "" {
+		return
+	}
+	helper.ResponseChunkData(c, streamResponse, data)
+}

relay/channel/openai/relay-openai.go

@@ -215,10 +215,35 @@ func OpenaiHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayI
 			StatusCode: resp.StatusCode,
 		}, nil
 	}
+	
+	forceFormat := false
+	if forceFmt, ok := info.ChannelSetting[constant.ForceFormat].(bool); ok {
+		forceFormat = forceFmt
+	}
+
+	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, info.UpstreamModelName)
+			completionTokens += ctkm
+		}
+		simpleResponse.Usage = dto.Usage{
+			PromptTokens:     info.PromptTokens,
+			CompletionTokens: completionTokens,
+			TotalTokens:      info.PromptTokens + completionTokens,
+		}
+	}
 
 	switch info.RelayFormat {
 	case relaycommon.RelayFormatOpenAI:
-		break
+		if forceFormat {
+			responseBody, err = json.Marshal(simpleResponse)
+			if err != nil {
+				return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil
+			}
+		} else {
+			break
+		}
 	case relaycommon.RelayFormatClaude:
 		claudeResp := service.ResponseOpenAI2Claude(&simpleResponse, info)
 		claudeRespStr, err := json.Marshal(claudeResp)
@@ -244,18 +269,6 @@ func OpenaiHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayI
 		common.SysError("error copying response body: " + err.Error())
 	}
 	resp.Body.Close()
-	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, info.UpstreamModelName)
-			completionTokens += ctkm
-		}
-		simpleResponse.Usage = dto.Usage{
-			PromptTokens:     info.PromptTokens,
-			CompletionTokens: completionTokens,
-			TotalTokens:      info.PromptTokens + completionTokens,
-		}
-	}
 	return nil, &simpleResponse.Usage
 }
 
@@ -595,3 +608,52 @@ func preConsumeUsage(ctx *gin.Context, info *relaycommon.RelayInfo, usage *dto.R
 	err := service.PreWssConsumeQuota(ctx, info, usage)
 	return err
 }
+
+func OpenaiHandlerWithUsage(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*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
+	}
+	// Reset response body
+	resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	// We shouldn't set the header before we parse the response body, because the parse part may fail.
+	// And then we will have to send an error response, but in this case, the header has already been set.
+	// So the httpClient will be confused by the response.
+	// For example, Postman will report error, and we cannot check the response at all.
+	for k, v := range resp.Header {
+		c.Writer.Header().Set(k, v[0])
+	}
+	// reset content length
+	c.Writer.Header().Set("Content-Length", fmt.Sprintf("%d", len(responseBody)))
+	c.Writer.WriteHeader(resp.StatusCode)
+	_, err = io.Copy(c.Writer, resp.Body)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "copy_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 usageResp dto.SimpleResponse
+	err = json.Unmarshal(responseBody, &usageResp)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "parse_response_body_failed", http.StatusInternalServerError), nil
+	}
+	// format
+	if usageResp.InputTokens > 0 {
+		usageResp.PromptTokens += usageResp.InputTokens
+	}
+	if usageResp.OutputTokens > 0 {
+		usageResp.CompletionTokens += usageResp.OutputTokens
+	}
+	if usageResp.InputTokensDetails != nil {
+		usageResp.PromptTokensDetails.ImageTokens += usageResp.InputTokensDetails.ImageTokens
+		usageResp.PromptTokensDetails.TextTokens += usageResp.InputTokensDetails.TextTokens
+	}
+	return nil, &usageResp.Usage
+}

relay/channel/openai/relay_responses.go

@@ -0,0 +1,119 @@
+package openai
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"net/http"
+	"one-api/common"
+	"one-api/dto"
+	relaycommon "one-api/relay/common"
+	"one-api/relay/helper"
+	"one-api/service"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+)
+
+func OaiResponsesHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	// read response body
+	var responsesResponse dto.OpenAIResponsesResponse
+	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
+	}
+	err = common.DecodeJson(responseBody, &responsesResponse)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if responsesResponse.Error != nil {
+		return &dto.OpenAIErrorWithStatusCode{
+			Error: dto.OpenAIError{
+				Message: responsesResponse.Error.Message,
+				Type:    "openai_error",
+				Code:    responsesResponse.Error.Code,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+
+	// reset response body
+	resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	// We shouldn't set the header before we parse the response body, because the parse part may fail.
+	// And then we will have to send an error response, but in this case, the header has already been set.
+	// So the httpClient will be confused by the response.
+	// For example, Postman will report error, and we cannot check the response at all.
+	for k, v := range resp.Header {
+		c.Writer.Header().Set(k, v[0])
+	}
+	c.Writer.WriteHeader(resp.StatusCode)
+	// copy response body
+	_, err = io.Copy(c.Writer, resp.Body)
+	if err != nil {
+		common.SysError("error copying response body: " + err.Error())
+	}
+	resp.Body.Close()
+	// compute usage
+	usage := dto.Usage{}
+	usage.PromptTokens = responsesResponse.Usage.InputTokens
+	usage.CompletionTokens = responsesResponse.Usage.OutputTokens
+	usage.TotalTokens = responsesResponse.Usage.TotalTokens
+	// 解析 Tools 用量
+	for _, tool := range responsesResponse.Tools {
+		info.ResponsesUsageInfo.BuiltInTools[tool.Type].CallCount++
+	}
+	return nil, &usage
+}
+
+func OaiResponsesStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	if resp == nil || resp.Body == nil {
+		common.LogError(c, "invalid response or response body")
+		return service.OpenAIErrorWrapper(fmt.Errorf("invalid response"), "invalid_response", http.StatusInternalServerError), nil
+	}
+
+	var usage = &dto.Usage{}
+	var responseTextBuilder strings.Builder
+
+	helper.StreamScannerHandler(c, resp, info, func(data string) bool {
+
+		// 检查当前数据是否包含 completed 状态和 usage 信息
+		var streamResponse dto.ResponsesStreamResponse
+		if err := common.DecodeJsonStr(data, &streamResponse); err == nil {
+			sendResponsesStreamData(c, streamResponse, data)
+			switch streamResponse.Type {
+			case "response.completed":
+				usage.PromptTokens = streamResponse.Response.Usage.InputTokens
+				usage.CompletionTokens = streamResponse.Response.Usage.OutputTokens
+				usage.TotalTokens = streamResponse.Response.Usage.TotalTokens
+			case "response.output_text.delta":
+				// 处理输出文本
+				responseTextBuilder.WriteString(streamResponse.Delta)
+			case dto.ResponsesOutputTypeItemDone:
+				// 函数调用处理
+				if streamResponse.Item != nil {
+					switch streamResponse.Item.Type {
+					case dto.BuildInCallWebSearchCall:
+						info.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview].CallCount++
+					}
+				}
+			}
+		}
+		return true
+	})
+
+	if usage.CompletionTokens == 0 {
+		// 计算输出文本的 token 数量
+		tempStr := responseTextBuilder.String()
+		if len(tempStr) > 0 {
+			// 非正常结束，使用输出文本的 token 数量
+			completionTokens, _ := service.CountTextToken(tempStr, info.UpstreamModelName)
+			usage.CompletionTokens = completionTokens
+		}
+	}
+
+	return nil, usage
+}

relay/channel/palm/adaptor.go

@@ -3,13 +3,14 @@ package palm
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
 	"one-api/relay/channel"
 	relaycommon "one-api/relay/common"
 	"one-api/service"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -60,6 +61,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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

@@ -3,13 +3,14 @@ package perplexity
 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"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -63,6 +64,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/siliconflow/adaptor.go

@@ -3,7 +3,6 @@ package siliconflow
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -11,6 +10,8 @@ import (
 	"one-api/relay/channel/openai"
 	relaycommon "one-api/relay/common"
 	"one-api/relay/constant"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -58,6 +59,11 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	return request, nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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)
 }
@@ -74,13 +80,9 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 	switch info.RelayMode {
 	case constant.RelayModeRerank:
 		err, usage = siliconflowRerankHandler(c, resp)
-	case constant.RelayModeChatCompletions:
-		if info.IsStream {
-			err, usage = openai.OaiStreamHandler(c, resp, info)
-		} else {
-			err, usage = openai.OpenaiHandler(c, resp, info)
-		}
 	case constant.RelayModeCompletions:
+		fallthrough
+	case constant.RelayModeChatCompletions:
 		if info.IsStream {
 			err, usage = openai.OaiStreamHandler(c, resp, info)
 		} else {

relay/channel/tencent/adaptor.go

@@ -3,7 +3,6 @@ package tencent
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/common"
@@ -13,6 +12,8 @@ import (
 	"one-api/service"
 	"strconv"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -84,6 +85,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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

@@ -4,7 +4,6 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -13,7 +12,10 @@ import (
 	"one-api/relay/channel/gemini"
 	"one-api/relay/channel/openai"
 	relaycommon "one-api/relay/common"
+	"one-api/setting/model_setting"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 const (
@@ -29,6 +31,8 @@ var claudeModelMap = map[string]string{
 	"claude-3-5-sonnet-20240620": "claude-3-5-sonnet@20240620",
 	"claude-3-5-sonnet-20241022": "claude-3-5-sonnet-v2@20241022",
 	"claude-3-7-sonnet-20250219": "claude-3-7-sonnet@20250219",
+	"claude-sonnet-4-20250514":   "claude-sonnet-4@20250514",
+	"claude-opus-4-20250514":     "claude-opus-4@20250514",
 }
 
 const anthropicVersion = "vertex-2023-10-16"
@@ -77,6 +81,15 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 	a.AccountCredentials = *adc
 	suffix := ""
 	if a.RequestMode == RequestModeGemini {
+		if model_setting.GetGeminiSettings().ThinkingAdapterEnabled {
+			// suffix -thinking and -nothinking
+			if strings.HasSuffix(info.OriginModelName, "-thinking") {
+				info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-thinking")
+			} else if strings.HasSuffix(info.OriginModelName, "-nothinking") {
+				info.UpstreamModelName = strings.TrimSuffix(info.UpstreamModelName, "-nothinking")
+			}
+		}
+
 		if info.IsStream {
 			suffix = "streamGenerateContent?alt=sse"
 		} else {
@@ -143,7 +156,7 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 		info.UpstreamModelName = claudeReq.Model
 		return vertexClaudeReq, nil
 	} else if a.RequestMode == RequestModeGemini {
-		geminiRequest, err := gemini.CovertGemini2OpenAI(*request)
+		geminiRequest, err := gemini.CovertGemini2OpenAI(*request, info)
 		if err != nil {
 			return nil, err
 		}
@@ -164,6 +177,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/volcengine/adaptor.go

@@ -3,7 +3,6 @@ package volcengine
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -12,6 +11,8 @@ import (
 	relaycommon "one-api/relay/common"
 	"one-api/relay/constant"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -71,6 +72,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return request, nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/xai/adaptor.go

@@ -2,14 +2,17 @@ package xai
 
 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"
 	"strings"
+
+	"one-api/relay/constant"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -27,15 +30,20 @@ func (a *Adaptor) ConvertAudioRequest(c *gin.Context, info *relaycommon.RelayInf
 }
 
 func (a *Adaptor) ConvertImageRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.ImageRequest) (any, error) {
-	request.Size = ""
-	return request, nil
+	xaiRequest := ImageRequest{
+		Model:          request.Model,
+		Prompt:         request.Prompt,
+		N:              request.N,
+		ResponseFormat: request.ResponseFormat,
+	}
+	return xaiRequest, nil
 }
 
 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
+	return relaycommon.GetFullRequestURL(info.BaseUrl, info.RequestURLPath, info.ChannelType), nil
 }
 
 func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Header, info *relaycommon.RelayInfo) error {
@@ -78,20 +86,26 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not available")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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)
 }
 
 func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) {
-	if info.IsStream {
-		err, usage = xAIStreamHandler(c, resp, info)
-	} else {
-		err, usage = xAIHandler(c, resp, info)
+	switch info.RelayMode {
+	case constant.RelayModeImagesGenerations, constant.RelayModeImagesEdits:
+		err, usage = openai.OpenaiHandlerWithUsage(c, resp, info)
+	default:
+		if info.IsStream {
+			err, usage = xAIStreamHandler(c, resp, info)
+		} else {
+			err, usage = xAIHandler(c, resp, info)
+		}
 	}
-	//if _, ok := usage.(*dto.Usage); ok && usage != nil {
-	//	usage.(*dto.Usage).CompletionTokens = usage.(*dto.Usage).TotalTokens - usage.(*dto.Usage).PromptTokens
-	//}
-
 	return
 }
 

relay/channel/xai/dto.go

@@ -12,3 +12,16 @@ type ChatCompletionResponse struct {
 	Usage             *dto.Usage `json:"usage"`
 	SystemFingerprint string     `json:"system_fingerprint"`
 }
+
+// quality, size or style are not supported by xAI API at the moment.
+type ImageRequest struct {
+	Model          string          `json:"model"`
+	Prompt         string          `json:"prompt" binding:"required"`
+	N              int             `json:"n,omitempty"`
+	// Size           string          `json:"size,omitempty"`
+	// Quality        string          `json:"quality,omitempty"`
+	ResponseFormat string          `json:"response_format,omitempty"`
+	// Style          string          `json:"style,omitempty"`
+	// User           string          `json:"user,omitempty"`
+	// ExtraFields    json.RawMessage `json:"extra_fields,omitempty"`
+}

relay/channel/xunfei/adaptor.go

@@ -2,7 +2,6 @@ package xunfei
 
 import (
 	"errors"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -10,6 +9,8 @@ import (
 	relaycommon "one-api/relay/common"
 	"one-api/service"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -61,6 +62,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return nil, errors.New("not implemented")
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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

@@ -3,12 +3,13 @@ package zhipu
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
 	"one-api/relay/channel"
 	relaycommon "one-api/relay/common"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -71,6 +72,11 @@ func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, request
 	return channel.DoApiRequest(a, c, info, requestBody)
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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 = zhipuStreamHandler(c, resp)

relay/channel/zhipu_4v/adaptor.go

@@ -3,7 +3,6 @@ package zhipu_4v
 import (
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/dto"
@@ -11,6 +10,8 @@ import (
 	"one-api/relay/channel/openai"
 	relaycommon "one-api/relay/common"
 	relayconstant "one-api/relay/constant"
+
+	"github.com/gin-gonic/gin"
 )
 
 type Adaptor struct {
@@ -70,6 +71,11 @@ func (a *Adaptor) ConvertEmbeddingRequest(c *gin.Context, info *relaycommon.Rela
 	return request, nil
 }
 
+func (a *Adaptor) ConvertOpenAIResponsesRequest(c *gin.Context, info *relaycommon.RelayInfo, request dto.OpenAIResponsesRequest) (any, error) {
+	// TODO implement me
+	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/common/relay_info.go

@@ -6,7 +6,6 @@ import (
 	"one-api/dto"
 	relayconstant "one-api/relay/constant"
 	"strings"
-	"sync"
 	"time"
 
 	"github.com/gin-gonic/gin"
@@ -37,6 +36,7 @@ type ClaudeConvertInfo struct {
 const (
 	RelayFormatOpenAI = "openai"
 	RelayFormatClaude = "claude"
+	RelayFormatGemini = "gemini"
 )
 
 type RerankerInfo struct {
@@ -44,6 +44,16 @@ type RerankerInfo struct {
 	ReturnDocuments bool
 }
 
+type BuildInToolInfo struct {
+	ToolName          string
+	CallCount         int
+	SearchContextSize string
+}
+
+type ResponsesUsageInfo struct {
+	BuiltInTools map[string]*BuildInToolInfo
+}
+
 type RelayInfo struct {
 	ChannelType       int
 	ChannelId         int
@@ -55,7 +65,6 @@ type RelayInfo struct {
 	StartTime         time.Time
 	FirstResponseTime time.Time
 	isFirstResponse   bool
-	responseMutex     sync.Mutex // Add mutex for protecting concurrent access
 	//SendLastReasoningResponse bool
 	ApiType           int
 	IsStream          bool
@@ -89,9 +98,11 @@ type RelayInfo struct {
 	UserQuota            int
 	RelayFormat          string
 	SendResponseCount    int
+	ChannelCreateTime    int64
 	ThinkingContentInfo
-	ClaudeConvertInfo
+	*ClaudeConvertInfo
 	*RerankerInfo
+	*ResponsesUsageInfo
 }
 
 // 定义支持流式选项的通道类型
@@ -105,6 +116,8 @@ var streamSupportedChannels = map[int]bool{
 	common.ChannelTypeVolcEngine: true,
 	common.ChannelTypeOllama:     true,
 	common.ChannelTypeXai:        true,
+	common.ChannelTypeDeepSeek:   true,
+	common.ChannelTypeBaiduV2:    true,
 }
 
 func GenRelayInfoWs(c *gin.Context, ws *websocket.Conn) *RelayInfo {
@@ -120,7 +133,7 @@ func GenRelayInfoClaude(c *gin.Context) *RelayInfo {
 	info := GenRelayInfo(c)
 	info.RelayFormat = RelayFormatClaude
 	info.ShouldIncludeUsage = false
-	info.ClaudeConvertInfo = ClaudeConvertInfo{
+	info.ClaudeConvertInfo = &ClaudeConvertInfo{
 		LastMessagesType: LastMessageTypeNone,
 	}
 	return info
@@ -136,6 +149,31 @@ func GenRelayInfoRerank(c *gin.Context, req *dto.RerankRequest) *RelayInfo {
 	return info
 }
 
+func GenRelayInfoResponses(c *gin.Context, req *dto.OpenAIResponsesRequest) *RelayInfo {
+	info := GenRelayInfo(c)
+	info.RelayMode = relayconstant.RelayModeResponses
+	info.ResponsesUsageInfo = &ResponsesUsageInfo{
+		BuiltInTools: make(map[string]*BuildInToolInfo),
+	}
+	if len(req.Tools) > 0 {
+		for _, tool := range req.Tools {
+			info.ResponsesUsageInfo.BuiltInTools[tool.Type] = &BuildInToolInfo{
+				ToolName:  tool.Type,
+				CallCount: 0,
+			}
+			switch tool.Type {
+			case dto.BuildInToolWebSearchPreview:
+				if tool.SearchContextSize == "" {
+					tool.SearchContextSize = "medium"
+				}
+				info.ResponsesUsageInfo.BuiltInTools[tool.Type].SearchContextSize = tool.SearchContextSize
+			}
+		}
+	}
+	info.IsStream = req.Stream
+	return info
+}
+
 func GenRelayInfo(c *gin.Context) *RelayInfo {
 	channelType := c.GetInt("channel_type")
 	channelId := c.GetInt("channel_id")
@@ -172,14 +210,15 @@ func GenRelayInfo(c *gin.Context) *RelayInfo {
 		OriginModelName:   c.GetString("original_model"),
 		UpstreamModelName: c.GetString("original_model"),
 		//RecodeModelName:   c.GetString("original_model"),
-		IsModelMapped:  false,
-		ApiType:        apiType,
-		ApiVersion:     c.GetString("api_version"),
-		ApiKey:         strings.TrimPrefix(c.Request.Header.Get("Authorization"), "Bearer "),
-		Organization:   c.GetString("channel_organization"),
-		ChannelSetting: channelSetting,
-		ParamOverride:  paramOverride,
-		RelayFormat:    RelayFormatOpenAI,
+		IsModelMapped:     false,
+		ApiType:           apiType,
+		ApiVersion:        c.GetString("api_version"),
+		ApiKey:            strings.TrimPrefix(c.Request.Header.Get("Authorization"), "Bearer "),
+		Organization:      c.GetString("channel_organization"),
+		ChannelSetting:    channelSetting,
+		ChannelCreateTime: c.GetInt64("channel_create_time"),
+		ParamOverride:     paramOverride,
+		RelayFormat:       RelayFormatOpenAI,
 		ThinkingContentInfo: ThinkingContentInfo{
 			IsFirstThinkingContent:  true,
 			SendLastThinkingContent: false,
@@ -202,6 +241,10 @@ func GenRelayInfo(c *gin.Context) *RelayInfo {
 	if streamSupportedChannels[info.ChannelType] {
 		info.SupportStreamOptions = true
 	}
+	// responses 模式不支持 StreamOptions
+	if relayconstant.RelayModeResponses == info.RelayMode {
+		info.SupportStreamOptions = false
+	}
 	return info
 }
 
@@ -214,9 +257,6 @@ func (info *RelayInfo) SetIsStream(isStream bool) {
 }
 
 func (info *RelayInfo) SetFirstResponseTime() {
-	info.responseMutex.Lock()
-	defer info.responseMutex.Unlock()
-
 	if info.isFirstResponse {
 		info.FirstResponseTime = time.Now()
 		info.isFirstResponse = false

relay/constant/api_type.go

@@ -33,6 +33,7 @@ const (
 	APITypeOpenRouter
 	APITypeXinference
 	APITypeXai
+	APITypeCoze
 	APITypeDummy // this one is only for count, do not add any channel after this
 )
 
@@ -95,6 +96,8 @@ func ChannelType2APIType(channelType int) (int, bool) {
 		apiType = APITypeXinference
 	case common.ChannelTypeXai:
 		apiType = APITypeXai
+	case common.ChannelTypeCoze:
+		apiType = APITypeCoze
 	}
 	if apiType == -1 {
 		return APITypeOpenAI, false

relay/constant/relay_mode.go

@@ -12,6 +12,7 @@ const (
 	RelayModeEmbeddings
 	RelayModeModerations
 	RelayModeImagesGenerations
+	RelayModeImagesEdits
 	RelayModeEdits
 
 	RelayModeMidjourneyImagine
@@ -39,6 +40,8 @@ const (
 
 	RelayModeRerank
 
+	RelayModeResponses
+
 	RelayModeRealtime
 )
 
@@ -56,8 +59,12 @@ func Path2RelayMode(path string) int {
 		relayMode = RelayModeModerations
 	} else if strings.HasPrefix(path, "/v1/images/generations") {
 		relayMode = RelayModeImagesGenerations
+	} else if strings.HasPrefix(path, "/v1/images/edits") {
+		relayMode = RelayModeImagesEdits
 	} else if strings.HasPrefix(path, "/v1/edits") {
 		relayMode = RelayModeEdits
+	} else if strings.HasPrefix(path, "/v1/responses") {
+		relayMode = RelayModeResponses
 	} else if strings.HasPrefix(path, "/v1/audio/speech") {
 		relayMode = RelayModeAudioSpeech
 	} else if strings.HasPrefix(path, "/v1/audio/transcriptions") {

relay/helper/common.go

@@ -12,11 +12,19 @@ import (
 )
 
 func SetEventStreamHeaders(c *gin.Context) {
-	c.Writer.Header().Set("Content-Type", "text/event-stream")
-	c.Writer.Header().Set("Cache-Control", "no-cache")
-	c.Writer.Header().Set("Connection", "keep-alive")
-	c.Writer.Header().Set("Transfer-Encoding", "chunked")
-	c.Writer.Header().Set("X-Accel-Buffering", "no")
+    // 检查是否已经设置过头部
+    if _, exists := c.Get("event_stream_headers_set"); exists {
+        return
+    }
+    
+    c.Writer.Header().Set("Content-Type", "text/event-stream")
+    c.Writer.Header().Set("Cache-Control", "no-cache")
+    c.Writer.Header().Set("Connection", "keep-alive")
+    c.Writer.Header().Set("Transfer-Encoding", "chunked")
+    c.Writer.Header().Set("X-Accel-Buffering", "no")
+    
+    // 设置标志，表示头部已经设置过
+    c.Set("event_stream_headers_set", true)
 }
 
 func ClaudeData(c *gin.Context, resp dto.ClaudeResponse) error {
@@ -43,6 +51,14 @@ func ClaudeChunkData(c *gin.Context, resp dto.ClaudeResponse, data string) {
 	}
 }
 
+func ResponseChunkData(c *gin.Context, resp dto.ResponsesStreamResponse, 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", 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/model_mapped.go

@@ -2,9 +2,11 @@ package helper
 
 import (
 	"encoding/json"
+	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"one-api/relay/common"
+
+	"github.com/gin-gonic/gin"
 )
 
 func ModelMappedHelper(c *gin.Context, info *common.RelayInfo) error {
@@ -16,9 +18,36 @@ func ModelMappedHelper(c *gin.Context, info *common.RelayInfo) error {
 		if err != nil {
 			return fmt.Errorf("unmarshal_model_mapping_failed")
 		}
-		if modelMap[info.OriginModelName] != "" {
-			info.UpstreamModelName = modelMap[info.OriginModelName]
-			info.IsModelMapped = true
+
+		// 支持链式模型重定向，最终使用链尾的模型
+		currentModel := info.OriginModelName
+		visitedModels := map[string]bool{
+			currentModel: true,
+		}
+		for {
+			if mappedModel, exists := modelMap[currentModel]; exists && mappedModel != "" {
+				// 模型重定向循环检测，避免无限循环
+				if visitedModels[mappedModel] {
+					if mappedModel == currentModel {
+						if currentModel == info.OriginModelName {
+							info.IsModelMapped = false
+							return nil
+						} else {
+							info.IsModelMapped = true
+							break
+						}
+					}
+					return errors.New("model_mapping_contains_cycle")
+				}
+				visitedModels[mappedModel] = true
+				currentModel = mappedModel
+				info.IsModelMapped = true
+			} else {
+				break
+			}
+		}
+		if info.IsModelMapped {
+			info.UpstreamModelName = currentModel
 		}
 	}
 	return nil

relay/helper/price.go

@@ -15,14 +15,15 @@ type PriceData struct {
 	ModelRatio             float64
 	CompletionRatio        float64
 	CacheRatio             float64
+	CacheCreationRatio     float64
+	ImageRatio             float64
 	GroupRatio             float64
 	UsePrice               bool
-	CacheCreationRatio     float64
 	ShouldPreConsumedQuota int
 }
 
 func (p PriceData) ToSetting() string {
-	return fmt.Sprintf("ModelPrice: %f, ModelRatio: %f, CompletionRatio: %f, CacheRatio: %f, GroupRatio: %f, UsePrice: %t, CacheCreationRatio: %f, ShouldPreConsumedQuota: %d", p.ModelPrice, p.ModelRatio, p.CompletionRatio, p.CacheRatio, p.GroupRatio, p.UsePrice, p.CacheCreationRatio, p.ShouldPreConsumedQuota)
+	return fmt.Sprintf("ModelPrice: %f, ModelRatio: %f, CompletionRatio: %f, CacheRatio: %f, GroupRatio: %f, UsePrice: %t, CacheCreationRatio: %f, ShouldPreConsumedQuota: %d, ImageRatio: %f", p.ModelPrice, p.ModelRatio, p.CompletionRatio, p.CacheRatio, p.GroupRatio, p.UsePrice, p.CacheCreationRatio, p.ShouldPreConsumedQuota, p.ImageRatio)
 }
 
 func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens int, maxTokens int) (PriceData, error) {
@@ -32,6 +33,7 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
 	var modelRatio float64
 	var completionRatio float64
 	var cacheRatio float64
+	var imageRatio float64
 	var cacheCreationRatio float64
 	if !usePrice {
 		preConsumedTokens := common.PreConsumedQuota
@@ -49,16 +51,13 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
 				}
 			}
 			if !acceptUnsetRatio {
-				if info.UserId == 1 {
-					return PriceData{}, fmt.Errorf("模型 %s 倍率或价格未配置，请设置或开始自用模式；Model %s ratio or price not set, please set or start self-use mode", info.OriginModelName, info.OriginModelName)
-				} else {
-					return PriceData{}, fmt.Errorf("模型 %s 倍率或价格未配置, 请联系管理员设置；Model %s ratio or price not set, please contact administrator to set", info.OriginModelName, info.OriginModelName)
-				}
+				return PriceData{}, fmt.Errorf("模型 %s 倍率或价格未配置，请联系管理员设置或开始自用模式；Model %s ratio or price not set, please set or start self-use mode", info.OriginModelName, info.OriginModelName)
 			}
 		}
 		completionRatio = operation_setting.GetCompletionRatio(info.OriginModelName)
 		cacheRatio, _ = operation_setting.GetCacheRatio(info.OriginModelName)
 		cacheCreationRatio, _ = operation_setting.GetCreateCacheRatio(info.OriginModelName)
+		imageRatio, _ = operation_setting.GetImageRatio(info.OriginModelName)
 		ratio := modelRatio * groupRatio
 		preConsumedQuota = int(float64(preConsumedTokens) * ratio)
 	} else {
@@ -72,6 +71,7 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
 		GroupRatio:             groupRatio,
 		UsePrice:               usePrice,
 		CacheRatio:             cacheRatio,
+		ImageRatio:             imageRatio,
 		CacheCreationRatio:     cacheCreationRatio,
 		ShouldPreConsumedQuota: preConsumedQuota,
 	}
@@ -82,3 +82,15 @@ func ModelPriceHelper(c *gin.Context, info *relaycommon.RelayInfo, promptTokens
 
 	return priceData, nil
 }
+
+func ContainPriceOrRatio(modelName string) bool {
+	_, ok := operation_setting.GetModelPrice(modelName, false)
+	if ok {
+		return true
+	}
+	_, ok = operation_setting.GetModelRatio(modelName)
+	if ok {
+		return true
+	}
+	return false
+}

relay/helper/stream_scanner.go

@@ -3,7 +3,6 @@ package helper
 import (
 	"bufio"
 	"context"
-	"github.com/bytedance/gopkg/util/gopool"
 	"io"
 	"net/http"
 	"one-api/common"
@@ -14,6 +13,8 @@ import (
 	"sync"
 	"time"
 
+	"github.com/bytedance/gopkg/util/gopool"
+
 	"github.com/gin-gonic/gin"
 )
 
@@ -32,7 +33,7 @@ func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon
 	defer resp.Body.Close()
 
 	streamingTimeout := time.Duration(constant.StreamingTimeout) * time.Second
-	if strings.HasPrefix(info.UpstreamModelName, "o1") || strings.HasPrefix(info.UpstreamModelName, "o3") {
+	if strings.HasPrefix(info.UpstreamModelName, "o") {
 		// twice timeout for thinking model
 		streamingTimeout *= 2
 	}
@@ -115,7 +116,7 @@ func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon
 			}
 			data = data[5:]
 			data = strings.TrimLeft(data, " ")
-			data = strings.TrimSuffix(data, "\"")
+			data = strings.TrimSuffix(data, "\r")
 			if !strings.HasPrefix(data, "[DONE]") {
 				info.SetFirstResponseTime()
 				writeMutex.Lock() // Lock before writing

relay/relay-image.go

@@ -5,21 +5,83 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"github.com/gin-gonic/gin"
 	"io"
 	"net/http"
 	"one-api/common"
 	"one-api/dto"
 	"one-api/model"
 	relaycommon "one-api/relay/common"
+	relayconstant "one-api/relay/constant"
 	"one-api/relay/helper"
 	"one-api/service"
 	"one-api/setting"
 	"strings"
+
+	"github.com/gin-gonic/gin"
 )
 
 func getAndValidImageRequest(c *gin.Context, info *relaycommon.RelayInfo) (*dto.ImageRequest, error) {
 	imageRequest := &dto.ImageRequest{}
+
+	switch info.RelayMode {
+	case relayconstant.RelayModeImagesEdits:
+		_, err := c.MultipartForm()
+		if err != nil {
+			return nil, err
+		}
+		formData := c.Request.PostForm
+		imageRequest.Prompt = formData.Get("prompt")
+		imageRequest.Model = formData.Get("model")
+		imageRequest.N = common.String2Int(formData.Get("n"))
+		imageRequest.Quality = formData.Get("quality")
+		imageRequest.Size = formData.Get("size")
+
+		if imageRequest.Model == "gpt-image-1" {
+			if imageRequest.Quality == "" {
+				imageRequest.Quality = "standard"
+			}
+		}
+	default:
+		err := common.UnmarshalBodyReusable(c, imageRequest)
+		if err != nil {
+			return nil, err
+		}
+		// Not "256x256", "512x512", or "1024x1024"
+		if imageRequest.Model == "dall-e-2" || imageRequest.Model == "dall-e" {
+			if imageRequest.Size != "" && imageRequest.Size != "256x256" && imageRequest.Size != "512x512" && imageRequest.Size != "1024x1024" {
+				return nil, errors.New("size must be one of 256x256, 512x512, or 1024x1024 for dall-e-2 or dall-e")
+			}
+		} else if imageRequest.Model == "dall-e-3" {
+			if imageRequest.Size != "" && imageRequest.Size != "1024x1024" && imageRequest.Size != "1024x1792" && imageRequest.Size != "1792x1024" {
+				return nil, errors.New("size must be one of 1024x1024, 1024x1792 or 1792x1024 for dall-e-3")
+			}
+			if imageRequest.Quality == "" {
+				imageRequest.Quality = "standard"
+			}
+			// N should between 1 and 10
+			//if imageRequest.N != 0 && (imageRequest.N < 1 || imageRequest.N > 10) {
+			//	return service.OpenAIErrorWrapper(errors.New("n must be between 1 and 10"), "invalid_field_value", http.StatusBadRequest)
+			//}
+		}
+	}
+
+	if imageRequest.Prompt == "" {
+		return nil, errors.New("prompt is required")
+	}
+
+	if imageRequest.Model == "" {
+		imageRequest.Model = "dall-e-2"
+	}
+	if strings.Contains(imageRequest.Size, "×") {
+		return nil, errors.New("size an unexpected error occurred in the parameter, please use 'x' instead of the multiplication sign '×'")
+	}
+	if imageRequest.N == 0 {
+		imageRequest.N = 1
+	}
+	if imageRequest.Size == "" {
+		imageRequest.Size = "1024x1024"
+	}
+
 	err := common.UnmarshalBodyReusable(c, imageRequest)
 	if err != nil {
 		return nil, err
@@ -39,6 +101,10 @@ func getAndValidImageRequest(c *gin.Context, info *relaycommon.RelayInfo) (*dto.
 	if imageRequest.Model == "" {
 		imageRequest.Model = "dall-e-2"
 	}
+	// x.ai grok-2-image not support size, quality or style
+	if imageRequest.Size == "empty" {
+		imageRequest.Size = ""
+	}
 
 	// Not "256x256", "512x512", or "1024x1024"
 	if imageRequest.Model == "dall-e-2" || imageRequest.Model == "dall-e" {
@@ -86,43 +152,59 @@ func ImageHelper(c *gin.Context) *dto.OpenAIErrorWithStatusCode {
 
 	imageRequest.Model = relayInfo.UpstreamModelName
 
-	priceData, err := helper.ModelPriceHelper(c, relayInfo, 0, 0)
+	priceData, err := helper.ModelPriceHelper(c, relayInfo, len(imageRequest.Prompt), 0)
 	if err != nil {
 		return service.OpenAIErrorWrapperLocal(err, "model_price_error", http.StatusInternalServerError)
 	}
+	var preConsumedQuota int
+	var quota int
+	var userQuota int
 	if !priceData.UsePrice {
 		// modelRatio 16 = modelPrice $0.04
 		// per 1 modelRatio = $0.04 / 16
-		priceData.ModelPrice = 0.0025 * priceData.ModelRatio
-	}
-
-	userQuota, err := model.GetUserQuota(relayInfo.UserId, false)
-
-	sizeRatio := 1.0
-	// Size
-	if imageRequest.Size == "256x256" {
-		sizeRatio = 0.4
-	} else if imageRequest.Size == "512x512" {
-		sizeRatio = 0.45
-	} else if imageRequest.Size == "1024x1024" {
-		sizeRatio = 1
-	} else if imageRequest.Size == "1024x1792" || imageRequest.Size == "1792x1024" {
-		sizeRatio = 2
-	}
-
-	qualityRatio := 1.0
-	if imageRequest.Model == "dall-e-3" && imageRequest.Quality == "hd" {
-		qualityRatio = 2.0
-		if imageRequest.Size == "1024x1792" || imageRequest.Size == "1792x1024" {
-			qualityRatio = 1.5
+		// priceData.ModelPrice = 0.0025 * priceData.ModelRatio
+		var openaiErr *dto.OpenAIErrorWithStatusCode
+		preConsumedQuota, userQuota, openaiErr = preConsumeQuota(c, priceData.ShouldPreConsumedQuota, relayInfo)
+		if openaiErr != nil {
+			return openaiErr
 		}
-	}
+		defer func() {
+			if openaiErr != nil {
+				returnPreConsumedQuota(c, relayInfo, userQuota, preConsumedQuota)
+			}
+		}()
 
-	priceData.ModelPrice *= sizeRatio * qualityRatio * float64(imageRequest.N)
-	quota := int(priceData.ModelPrice * priceData.GroupRatio * common.QuotaPerUnit)
+	} else {
+		sizeRatio := 1.0
+		// Size
+		if imageRequest.Size == "256x256" {
+			sizeRatio = 0.4
+		} else if imageRequest.Size == "512x512" {
+			sizeRatio = 0.45
+		} else if imageRequest.Size == "1024x1024" {
+			sizeRatio = 1
+		} else if imageRequest.Size == "1024x1792" || imageRequest.Size == "1792x1024" {
+			sizeRatio = 2
+		}
 
-	if userQuota-quota < 0 {
-		return service.OpenAIErrorWrapperLocal(fmt.Errorf("image pre-consumed quota failed, user quota: %s, need quota: %s", common.FormatQuota(userQuota), common.FormatQuota(quota)), "insufficient_user_quota", http.StatusForbidden)
+		qualityRatio := 1.0
+		if imageRequest.Model == "dall-e-3" && imageRequest.Quality == "hd" {
+			qualityRatio = 2.0
+			if imageRequest.Size == "1024x1792" || imageRequest.Size == "1792x1024" {
+				qualityRatio = 1.5
+			}
+		}
+
+		// reset model price
+		priceData.ModelPrice *= sizeRatio * qualityRatio * float64(imageRequest.N)
+		quota = int(priceData.ModelPrice * priceData.GroupRatio * common.QuotaPerUnit)
+		userQuota, err = model.GetUserQuota(relayInfo.UserId, false)
+		if err != nil {
+			return service.OpenAIErrorWrapperLocal(err, "get_user_quota_failed", http.StatusInternalServerError)
+		}
+		if userQuota-quota < 0 {
+			return service.OpenAIErrorWrapperLocal(fmt.Errorf("image pre-consumed quota failed, user quota: %s, need quota: %s", common.FormatQuota(userQuota), common.FormatQuota(quota)), "insufficient_user_quota", http.StatusForbidden)
+		}
 	}
 
 	adaptor := GetAdaptor(relayInfo.ApiType)
@@ -137,12 +219,15 @@ func ImageHelper(c *gin.Context) *dto.OpenAIErrorWithStatusCode {
 	if err != nil {
 		return service.OpenAIErrorWrapperLocal(err, "convert_request_failed", http.StatusInternalServerError)
 	}
-
-	jsonData, err := json.Marshal(convertedRequest)
-	if err != nil {
-		return service.OpenAIErrorWrapperLocal(err, "json_marshal_failed", http.StatusInternalServerError)
+	if relayInfo.RelayMode == relayconstant.RelayModeImagesEdits {
+		requestBody = convertedRequest.(io.Reader)
+	} else {
+		jsonData, err := json.Marshal(convertedRequest)
+		if err != nil {
+			return service.OpenAIErrorWrapperLocal(err, "json_marshal_failed", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(jsonData)
 	}
-	requestBody = bytes.NewBuffer(jsonData)
 
 	statusCodeMappingStr := c.GetString("status_code_mapping")
 
@@ -162,24 +247,25 @@ func ImageHelper(c *gin.Context) *dto.OpenAIErrorWithStatusCode {
 		}
 	}
 
-	_, openaiErr := adaptor.DoResponse(c, httpResp, relayInfo)
+	usage, openaiErr := adaptor.DoResponse(c, httpResp, relayInfo)
 	if openaiErr != nil {
 		// reset status code 重置状态码
 		service.ResetStatusCode(openaiErr, statusCodeMappingStr)
 		return openaiErr
 	}
 
-	usage := &dto.Usage{
-		PromptTokens: imageRequest.N,
-		TotalTokens:  imageRequest.N,
+	if usage.(*dto.Usage).TotalTokens == 0 {
+		usage.(*dto.Usage).TotalTokens = imageRequest.N
+	}
+	if usage.(*dto.Usage).PromptTokens == 0 {
+		usage.(*dto.Usage).PromptTokens = imageRequest.N
 	}
-
 	quality := "standard"
 	if imageRequest.Quality == "hd" {
 		quality = "hd"
 	}
 
 	logContent := fmt.Sprintf("大小 %s, 品质 %s", imageRequest.Size, quality)
-	postConsumeQuota(c, relayInfo, usage, 0, userQuota, priceData, logContent)
+	postConsumeQuota(c, relayInfo, usage.(*dto.Usage), preConsumedQuota, userQuota, priceData, logContent)
 	return nil
 }

relay/relay-mj.go

@@ -32,7 +32,23 @@ func RelayMidjourneyImage(c *gin.Context) {
 		})
 		return
 	}
-	resp, err := http.Get(midjourneyTask.ImageUrl)
+	var httpClient *http.Client
+	if channel, err := model.CacheGetChannel(midjourneyTask.ChannelId); err == nil {
+		if proxy, ok := channel.GetSetting()["proxy"]; ok {
+			if proxyURL, ok := proxy.(string); ok && proxyURL != "" {
+				if httpClient, err = service.NewProxyHttpClient(proxyURL); err != nil {
+					c.JSON(400, gin.H{
+						"error": "proxy_url_invalid",
+					})
+					return
+				}
+			}
+		}
+	}
+	if httpClient == nil {
+		httpClient = service.GetHttpClient()
+	}
+	resp, err := httpClient.Get(midjourneyTask.ImageUrl)
 	if err != nil {
 		c.JSON(http.StatusInternalServerError, gin.H{
 			"error": "http_get_image_failed",

relay/relay-responses.go

@@ -0,0 +1,171 @@
+package relay
+
+import (
+	"bytes"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"one-api/common"
+	"one-api/dto"
+	relaycommon "one-api/relay/common"
+	"one-api/relay/helper"
+	"one-api/service"
+	"one-api/setting"
+	"one-api/setting/model_setting"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+)
+
+func getAndValidateResponsesRequest(c *gin.Context) (*dto.OpenAIResponsesRequest, error) {
+	request := &dto.OpenAIResponsesRequest{}
+	err := common.UnmarshalBodyReusable(c, request)
+	if err != nil {
+		return nil, err
+	}
+	if request.Model == "" {
+		return nil, errors.New("model is required")
+	}
+	if len(request.Input) == 0 {
+		return nil, errors.New("input is required")
+	}
+	return request, nil
+
+}
+
+func checkInputSensitive(textRequest *dto.OpenAIResponsesRequest, info *relaycommon.RelayInfo) ([]string, error) {
+	sensitiveWords, err := service.CheckSensitiveInput(textRequest.Input)
+	return sensitiveWords, err
+}
+
+func getInputTokens(req *dto.OpenAIResponsesRequest, info *relaycommon.RelayInfo) (int, error) {
+	inputTokens, err := service.CountTokenInput(req.Input, req.Model)
+	info.PromptTokens = inputTokens
+	return inputTokens, err
+}
+
+func ResponsesHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) {
+	req, err := getAndValidateResponsesRequest(c)
+	if err != nil {
+		common.LogError(c, fmt.Sprintf("getAndValidateResponsesRequest error: %s", err.Error()))
+		return service.OpenAIErrorWrapperLocal(err, "invalid_responses_request", http.StatusBadRequest)
+	}
+
+	relayInfo := relaycommon.GenRelayInfoResponses(c, req)
+
+	if setting.ShouldCheckPromptSensitive() {
+		sensitiveWords, err := checkInputSensitive(req, relayInfo)
+		if err != nil {
+			common.LogWarn(c, fmt.Sprintf("user sensitive words detected: %s", strings.Join(sensitiveWords, ", ")))
+			return service.OpenAIErrorWrapperLocal(err, "check_request_sensitive_error", http.StatusBadRequest)
+		}
+	}
+
+	err = helper.ModelMappedHelper(c, relayInfo)
+	if err != nil {
+		return service.OpenAIErrorWrapperLocal(err, "model_mapped_error", http.StatusBadRequest)
+	}
+	req.Model = relayInfo.UpstreamModelName
+	if value, exists := c.Get("prompt_tokens"); exists {
+		promptTokens := value.(int)
+		relayInfo.SetPromptTokens(promptTokens)
+	} else {
+		promptTokens, err := getInputTokens(req, relayInfo)
+		if err != nil {
+			return service.OpenAIErrorWrapperLocal(err, "count_input_tokens_error", http.StatusBadRequest)
+		}
+		c.Set("prompt_tokens", promptTokens)
+	}
+
+	priceData, err := helper.ModelPriceHelper(c, relayInfo, relayInfo.PromptTokens, int(req.MaxOutputTokens))
+	if err != nil {
+		return service.OpenAIErrorWrapperLocal(err, "model_price_error", http.StatusInternalServerError)
+	}
+	// pre consume quota
+	preConsumedQuota, userQuota, openaiErr := preConsumeQuota(c, priceData.ShouldPreConsumedQuota, relayInfo)
+	if openaiErr != nil {
+		return openaiErr
+	}
+	defer func() {
+		if openaiErr != nil {
+			returnPreConsumedQuota(c, relayInfo, userQuota, preConsumedQuota)
+		}
+	}()
+	adaptor := GetAdaptor(relayInfo.ApiType)
+	if adaptor == nil {
+		return service.OpenAIErrorWrapperLocal(fmt.Errorf("invalid api type: %d", relayInfo.ApiType), "invalid_api_type", http.StatusBadRequest)
+	}
+	adaptor.Init(relayInfo)
+	var requestBody io.Reader
+	if model_setting.GetGlobalSettings().PassThroughRequestEnabled {
+		body, err := common.GetRequestBody(c)
+		if err != nil {
+			return service.OpenAIErrorWrapperLocal(err, "get_request_body_error", http.StatusInternalServerError)
+		}
+		requestBody = bytes.NewBuffer(body)
+	} else {
+		convertedRequest, err := adaptor.ConvertOpenAIResponsesRequest(c, relayInfo, *req)
+		if err != nil {
+			return service.OpenAIErrorWrapperLocal(err, "convert_request_error", http.StatusBadRequest)
+		}
+		jsonData, err := json.Marshal(convertedRequest)
+		if err != nil {
+			return service.OpenAIErrorWrapperLocal(err, "marshal_request_error", http.StatusInternalServerError)
+		}
+		// apply param override
+		if len(relayInfo.ParamOverride) > 0 {
+			reqMap := make(map[string]interface{})
+			err = json.Unmarshal(jsonData, &reqMap)
+			if err != nil {
+				return service.OpenAIErrorWrapperLocal(err, "param_override_unmarshal_failed", http.StatusInternalServerError)
+			}
+			for key, value := range relayInfo.ParamOverride {
+				reqMap[key] = value
+			}
+			jsonData, err = json.Marshal(reqMap)
+			if err != nil {
+				return service.OpenAIErrorWrapperLocal(err, "param_override_marshal_failed", http.StatusInternalServerError)
+			}
+		}
+
+		if common.DebugEnabled {
+			println("requestBody: ", string(jsonData))
+		}
+		requestBody = bytes.NewBuffer(jsonData)
+	}
+
+	var httpResp *http.Response
+	resp, err := adaptor.DoRequest(c, relayInfo, requestBody)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "do_request_failed", http.StatusInternalServerError)
+	}
+
+	statusCodeMappingStr := c.GetString("status_code_mapping")
+
+	if resp != nil {
+		httpResp = resp.(*http.Response)
+
+		if httpResp.StatusCode != http.StatusOK {
+			openaiErr = service.RelayErrorHandler(httpResp, false)
+			// reset status code 重置状态码
+			service.ResetStatusCode(openaiErr, statusCodeMappingStr)
+			return openaiErr
+		}
+	}
+
+	usage, openaiErr := adaptor.DoResponse(c, httpResp, relayInfo)
+	if openaiErr != nil {
+		// reset status code 重置状态码
+		service.ResetStatusCode(openaiErr, statusCodeMappingStr)
+		return openaiErr
+	}
+
+	if strings.HasPrefix(relayInfo.OriginModelName, "gpt-4o-audio") {
+		service.PostAudioConsumeQuota(c, relayInfo, usage.(*dto.Usage), preConsumedQuota, userQuota, priceData, "")
+	} else {
+		postConsumeQuota(c, relayInfo, usage.(*dto.Usage), preConsumedQuota, userQuota, priceData, "")
+	}
+	return nil
+}

relay/relay-text.go

@@ -18,6 +18,7 @@ import (
 	"one-api/service"
 	"one-api/setting"
 	"one-api/setting/model_setting"
+	"one-api/setting/operation_setting"
 	"strings"
 	"time"
 
@@ -193,6 +194,7 @@ func TextHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) {
 
 	var httpResp *http.Response
 	resp, err := adaptor.DoRequest(c, relayInfo, requestBody)
+
 	if err != nil {
 		return service.OpenAIErrorWrapper(err, "do_request_failed", http.StatusInternalServerError)
 	}
@@ -331,12 +333,14 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 	useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
 	promptTokens := usage.PromptTokens
 	cacheTokens := usage.PromptTokensDetails.CachedTokens
+	imageTokens := usage.PromptTokensDetails.ImageTokens
 	completionTokens := usage.CompletionTokens
 	modelName := relayInfo.OriginModelName
 
 	tokenName := ctx.GetString("token_name")
 	completionRatio := priceData.CompletionRatio
 	cacheRatio := priceData.CacheRatio
+	imageRatio := priceData.ImageRatio
 	modelRatio := priceData.ModelRatio
 	groupRatio := priceData.GroupRatio
 	modelPrice := priceData.ModelPrice
@@ -344,9 +348,11 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 	// Convert values to decimal for precise calculation
 	dPromptTokens := decimal.NewFromInt(int64(promptTokens))
 	dCacheTokens := decimal.NewFromInt(int64(cacheTokens))
+	dImageTokens := decimal.NewFromInt(int64(imageTokens))
 	dCompletionTokens := decimal.NewFromInt(int64(completionTokens))
 	dCompletionRatio := decimal.NewFromFloat(completionRatio)
 	dCacheRatio := decimal.NewFromFloat(cacheRatio)
+	dImageRatio := decimal.NewFromFloat(imageRatio)
 	dModelRatio := decimal.NewFromFloat(modelRatio)
 	dGroupRatio := decimal.NewFromFloat(groupRatio)
 	dModelPrice := decimal.NewFromFloat(modelPrice)
@@ -354,11 +360,46 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 
 	ratio := dModelRatio.Mul(dGroupRatio)
 
+	// openai web search 工具计费
+	var dWebSearchQuota decimal.Decimal
+	var webSearchPrice float64
+	if relayInfo.ResponsesUsageInfo != nil {
+		if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists && webSearchTool.CallCount > 0 {
+			// 计算 web search 调用的配额 (配额 = 价格 * 调用次数 / 1000 * 分组倍率)
+			webSearchPrice = operation_setting.GetWebSearchPricePerThousand(modelName, webSearchTool.SearchContextSize)
+			dWebSearchQuota = decimal.NewFromFloat(webSearchPrice).
+				Mul(decimal.NewFromInt(int64(webSearchTool.CallCount))).
+				Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+			extraContent += fmt.Sprintf("Web Search 调用 %d 次，上下文大小 %s，调用花费 $%s",
+				webSearchTool.CallCount, webSearchTool.SearchContextSize, dWebSearchQuota.String())
+		}
+	}
+	// file search tool 计费
+	var dFileSearchQuota decimal.Decimal
+	var fileSearchPrice float64
+	if relayInfo.ResponsesUsageInfo != nil {
+		if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists && fileSearchTool.CallCount > 0 {
+			fileSearchPrice = operation_setting.GetFileSearchPricePerThousand()
+			dFileSearchQuota = decimal.NewFromFloat(fileSearchPrice).
+				Mul(decimal.NewFromInt(int64(fileSearchTool.CallCount))).
+				Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+			extraContent += fmt.Sprintf("File Search 调用 %d 次，调用花费 $%s",
+				fileSearchTool.CallCount, dFileSearchQuota.String())
+		}
+	}
+
 	var quotaCalculateDecimal decimal.Decimal
 	if !priceData.UsePrice {
 		nonCachedTokens := dPromptTokens.Sub(dCacheTokens)
 		cachedTokensWithRatio := dCacheTokens.Mul(dCacheRatio)
+
 		promptQuota := nonCachedTokens.Add(cachedTokensWithRatio)
+		if imageTokens > 0 {
+			nonImageTokens := dPromptTokens.Sub(dImageTokens)
+			imageTokensWithRatio := dImageTokens.Mul(dImageRatio)
+			promptQuota = nonImageTokens.Add(imageTokensWithRatio)
+		}
+
 		completionQuota := dCompletionTokens.Mul(dCompletionRatio)
 
 		quotaCalculateDecimal = promptQuota.Add(completionQuota).Mul(ratio)
@@ -369,6 +410,9 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 	} else {
 		quotaCalculateDecimal = dModelPrice.Mul(dQuotaPerUnit).Mul(dGroupRatio)
 	}
+	// 添加 responses tools call 调用的配额
+	quotaCalculateDecimal = quotaCalculateDecimal.Add(dWebSearchQuota)
+	quotaCalculateDecimal = quotaCalculateDecimal.Add(dFileSearchQuota)
 
 	quota := int(quotaCalculateDecimal.Round(0).IntPart())
 	totalTokens := promptTokens + completionTokens
@@ -414,6 +458,25 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 		logContent += ", " + extraContent
 	}
 	other := service.GenerateTextOtherInfo(ctx, relayInfo, modelRatio, groupRatio, completionRatio, cacheTokens, cacheRatio, modelPrice)
+	if imageTokens != 0 {
+		other["image"] = true
+		other["image_ratio"] = imageRatio
+		other["image_output"] = imageTokens
+	}
+	if !dWebSearchQuota.IsZero() && relayInfo.ResponsesUsageInfo != nil {
+		if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists {
+			other["web_search"] = true
+			other["web_search_call_count"] = webSearchTool.CallCount
+			other["web_search_price"] = webSearchPrice
+		}
+	}
+	if !dFileSearchQuota.IsZero() && relayInfo.ResponsesUsageInfo != nil {
+		if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists {
+			other["file_search"] = true
+			other["file_search_call_count"] = fileSearchTool.CallCount
+			other["file_search_price"] = fileSearchPrice
+		}
+	}
 	model.RecordConsumeLog(ctx, relayInfo.UserId, relayInfo.ChannelId, promptTokens, completionTokens, logModel,
 		tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, relayInfo.Group, other)
 }

relay/relay_adaptor.go

@@ -10,6 +10,7 @@ import (
 	"one-api/relay/channel/claude"
 	"one-api/relay/channel/cloudflare"
 	"one-api/relay/channel/cohere"
+	"one-api/relay/channel/coze"
 	"one-api/relay/channel/deepseek"
 	"one-api/relay/channel/dify"
 	"one-api/relay/channel/gemini"
@@ -88,6 +89,8 @@ func GetAdaptor(apiType int) channel.Adaptor {
 		return &openai.Adaptor{}
 	case constant.APITypeXai:
 		return &xai.Adaptor{}
+	case constant.APITypeCoze:
+		return &coze.Adaptor{}
 	}
 	return nil
 }

router/relay-router.go

@@ -1,10 +1,11 @@
 package router
 
 import (
-	"github.com/gin-gonic/gin"
 	"one-api/controller"
 	"one-api/middleware"
 	"one-api/relay"
+
+	"github.com/gin-gonic/gin"
 )
 
 func SetRelayRouter(router *gin.Engine) {
@@ -40,13 +41,14 @@ func SetRelayRouter(router *gin.Engine) {
 		httpRouter.POST("/chat/completions", controller.Relay)
 		httpRouter.POST("/edits", controller.Relay)
 		httpRouter.POST("/images/generations", controller.Relay)
-		httpRouter.POST("/images/edits", controller.RelayNotImplemented)
+		httpRouter.POST("/images/edits", controller.Relay)
 		httpRouter.POST("/images/variations", controller.RelayNotImplemented)
 		httpRouter.POST("/embeddings", controller.Relay)
 		httpRouter.POST("/engines/:model/embeddings", controller.Relay)
 		httpRouter.POST("/audio/transcriptions", controller.Relay)
 		httpRouter.POST("/audio/translations", controller.Relay)
 		httpRouter.POST("/audio/speech", controller.Relay)
+		httpRouter.POST("/responses", controller.Relay)
 		httpRouter.GET("/files", controller.RelayNotImplemented)
 		httpRouter.POST("/files", controller.RelayNotImplemented)
 		httpRouter.DELETE("/files/:id", controller.RelayNotImplemented)

service/cf_worker.go

@@ -24,7 +24,7 @@ func DoWorkerRequest(req *WorkerRequest) (*http.Response, error) {
 	if !setting.EnableWorker() {
 		return nil, fmt.Errorf("worker not enabled")
 	}
-	if !strings.HasPrefix(req.URL, "https") {
+	if !setting.WorkerAllowHttpImageRequestEnabled && !strings.HasPrefix(req.URL, "https") {
 		return nil, fmt.Errorf("only support https url")
 	}
 

service/http_client.go

@@ -3,12 +3,13 @@ package service
 import (
 	"context"
 	"fmt"
-	"golang.org/x/net/proxy"
 	"net"
 	"net/http"
 	"net/url"
 	"one-api/common"
 	"time"
+
+	"golang.org/x/net/proxy"
 )
 
 var httpClient *http.Client
@@ -55,7 +56,7 @@ func NewProxyHttpClient(proxyURL string) (*http.Client, error) {
 			},
 		}, nil
 
-	case "socks5":
+	case "socks5", "socks5h":
 		// 获取认证信息
 		var auth *proxy.Auth
 		if parsedURL.User != nil {
@@ -69,6 +70,7 @@ func NewProxyHttpClient(proxyURL string) (*http.Client, error) {
 		}
 
 		// 创建 SOCKS5 代理拨号器
+		// proxy.SOCKS5 使用 tcp 参数，所有 TCP 连接包括 DNS 查询都将通过代理进行。行为与 socks5h 相同
 		dialer, err := proxy.SOCKS5("tcp", parsedURL.Host, auth, proxy.Direct)
 		if err != nil {
 			return nil, err

service/token_counter.go

@@ -43,7 +43,7 @@ func InitTokenEncoders() {
 			} else {
 				tokenEncoderMap[model] = defaultTokenEncoder
 			}
-		} else if strings.HasPrefix(model, "o1") {
+		} else if strings.HasPrefix(model, "o") {
 			tokenEncoderMap[model] = o200kTokenEncoder
 		} else {
 			tokenEncoderMap[model] = defaultTokenEncoder
@@ -120,11 +120,12 @@ func getImageToken(info *relaycommon.RelayInfo, imageUrl *dto.MessageImageUrl, m
 	var config image.Config
 	var err error
 	var format string
+	var b64str string
 	if strings.HasPrefix(imageUrl.Url, "http") {
 		config, format, err = DecodeUrlImageData(imageUrl.Url)
 	} else {
 		common.SysLog(fmt.Sprintf("decoding image"))
-		config, format, _, err = DecodeBase64ImageData(imageUrl.Url)
+		config, format, b64str, err = DecodeBase64ImageData(imageUrl.Url)
 	}
 	if err != nil {
 		return 0, err
@@ -132,7 +133,12 @@ func getImageToken(info *relaycommon.RelayInfo, imageUrl *dto.MessageImageUrl, m
 	imageUrl.MimeType = format
 
 	if config.Width == 0 || config.Height == 0 {
-		return 0, errors.New(fmt.Sprintf("fail to decode image config: %s", imageUrl.Url))
+		// not an image
+		if format != "" && b64str != "" {
+			// file type
+			return 3 * baseTokens, nil
+		}
+		return 0, errors.New(fmt.Sprintf("fail to decode base64 config: %s", imageUrl.Url))
 	}
 
 	shortSide := config.Width
@@ -400,6 +406,8 @@ func CountTokenMessages(info *relaycommon.RelayInfo, messages []dto.Message, mod
 					tokenNum += 100
 				} else if m.Type == dto.ContentTypeFile {
 					tokenNum += 5000
+				} else if m.Type == dto.ContentTypeVideoUrl {
+					tokenNum += 5000
 				} else {
 					tokenNum += getTokenNum(tokenEncoder, m.Text)
 				}

setting/model_setting/gemini.go

@@ -6,8 +6,11 @@ import (
 
 // GeminiSettings 定义Gemini模型的配置
 type GeminiSettings struct {
-	SafetySettings  map[string]string `json:"safety_settings"`
-	VersionSettings map[string]string `json:"version_settings"`
+	SafetySettings                        map[string]string `json:"safety_settings"`
+	VersionSettings                       map[string]string `json:"version_settings"`
+	SupportedImagineModels                []string          `json:"supported_imagine_models"`
+	ThinkingAdapterEnabled                bool              `json:"thinking_adapter_enabled"`
+	ThinkingAdapterBudgetTokensPercentage float64           `json:"thinking_adapter_budget_tokens_percentage"`
 }
 
 // 默认配置
@@ -20,6 +23,12 @@ var defaultGeminiSettings = GeminiSettings{
 		"default":        "v1beta",
 		"gemini-1.0-pro": "v1",
 	},
+	SupportedImagineModels: []string{
+		"gemini-2.0-flash-exp-image-generation",
+		"gemini-2.0-flash-exp",
+	},
+	ThinkingAdapterEnabled:                false,
+	ThinkingAdapterBudgetTokensPercentage: 0.6,
 }
 
 // 全局实例
@@ -50,3 +59,12 @@ func GetGeminiVersionSetting(key string) string {
 	}
 	return geminiSettings.VersionSettings["default"]
 }
+
+func IsGeminiModelSupportImagine(model string) bool {
+	for _, v := range geminiSettings.SupportedImagineModels {
+		if v == model {
+			return true
+		}
+	}
+	return false
+}

setting/operation_setting/cache_ratio.go

@@ -36,6 +36,10 @@ var defaultCacheRatio = map[string]float64{
 	"claude-3-5-sonnet-20241022":          0.1,
 	"claude-3-7-sonnet-20250219":          0.1,
 	"claude-3-7-sonnet-20250219-thinking": 0.1,
+	"claude-sonnet-4-20250514":            0.1,
+	"claude-sonnet-4-20250514-thinking":   0.1,
+	"claude-opus-4-20250514":              0.1,
+	"claude-opus-4-20250514-thinking":     0.1,
 }
 
 var defaultCreateCacheRatio = map[string]float64{
@@ -47,6 +51,10 @@ var defaultCreateCacheRatio = map[string]float64{
 	"claude-3-5-sonnet-20241022":          1.25,
 	"claude-3-7-sonnet-20250219":          1.25,
 	"claude-3-7-sonnet-20250219-thinking": 1.25,
+	"claude-sonnet-4-20250514":            1.25,
+	"claude-sonnet-4-20250514-thinking":   1.25,
+	"claude-opus-4-20250514":              1.25,
+	"claude-opus-4-20250514-thinking":     1.25,
 }
 
 //var defaultCreateCacheRatio = map[string]float64{}

setting/operation_setting/model-ratio.go

@@ -51,26 +51,27 @@ var defaultModelRatio = map[string]float64{
 	"gpt-4o-realtime-preview-2024-12-17":      2.5,
 	"gpt-4o-mini-realtime-preview":            0.3,
 	"gpt-4o-mini-realtime-preview-2024-12-17": 0.3,
-	"o1":                         7.5,
-	"o1-2024-12-17":              7.5,
-	"o1-preview":                 7.5,
-	"o1-preview-2024-09-12":      7.5,
-	"o1-mini":                    0.55,
-	"o1-mini-2024-09-12":         0.55,
-	"o3-mini":                    0.55,
-	"o3-mini-2025-01-31":         0.55,
-	"o3-mini-high":               0.55,
-	"o3-mini-2025-01-31-high":    0.55,
-	"o3-mini-low":                0.55,
-	"o3-mini-2025-01-31-low":     0.55,
-	"o3-mini-medium":             0.55,
-	"o3-mini-2025-01-31-medium":  0.55,
-	"gpt-4o-mini":                0.075,
-	"gpt-4o-mini-2024-07-18":     0.075,
-	"gpt-4-turbo":                5, // $0.01 / 1K tokens
-	"gpt-4-turbo-2024-04-09":     5, // $0.01 / 1K tokens
-	"gpt-4.5-preview":            37.5,
-	"gpt-4.5-preview-2025-02-27": 37.5,
+	"gpt-image-1":                             2.5,
+	"o1":                                      7.5,
+	"o1-2024-12-17":                           7.5,
+	"o1-preview":                              7.5,
+	"o1-preview-2024-09-12":                   7.5,
+	"o1-mini":                                 0.55,
+	"o1-mini-2024-09-12":                      0.55,
+	"o3-mini":                                 0.55,
+	"o3-mini-2025-01-31":                      0.55,
+	"o3-mini-high":                            0.55,
+	"o3-mini-2025-01-31-high":                 0.55,
+	"o3-mini-low":                             0.55,
+	"o3-mini-2025-01-31-low":                  0.55,
+	"o3-mini-medium":                          0.55,
+	"o3-mini-2025-01-31-medium":               0.55,
+	"gpt-4o-mini":                             0.075,
+	"gpt-4o-mini-2024-07-18":                  0.075,
+	"gpt-4-turbo":                             5, // $0.01 / 1K tokens
+	"gpt-4-turbo-2024-04-09":                  5, // $0.01 / 1K tokens
+	"gpt-4.5-preview":                         37.5,
+	"gpt-4.5-preview-2025-02-27":              37.5,
 	//"gpt-3.5-turbo-0301":           0.75, //deprecated
 	"gpt-3.5-turbo":          0.25,
 	"gpt-3.5-turbo-0613":     0.75,
@@ -86,89 +87,94 @@ var defaultModelRatio = map[string]float64{
 	"text-curie-001":         1,
 	//"text-davinci-002":               10,
 	//"text-davinci-003":               10,
-	"text-davinci-edit-001":               10,
-	"code-davinci-edit-001":               10,
-	"whisper-1":                           15,  // $0.006 / minute -> $0.006 / 150 words -> $0.006 / 200 tokens -> $0.03 / 1k tokens
-	"tts-1":                               7.5, // 1k characters -> $0.015
-	"tts-1-1106":                          7.5, // 1k characters -> $0.015
-	"tts-1-hd":                            15,  // 1k characters -> $0.03
-	"tts-1-hd-1106":                       15,  // 1k characters -> $0.03
-	"davinci":                             10,
-	"curie":                               10,
-	"babbage":                             10,
-	"ada":                                 10,
-	"text-embedding-3-small":              0.01,
-	"text-embedding-3-large":              0.065,
-	"text-embedding-ada-002":              0.05,
-	"text-search-ada-doc-001":             10,
-	"text-moderation-stable":              0.1,
-	"text-moderation-latest":              0.1,
-	"claude-instant-1":                    0.4,   // $0.8 / 1M tokens
-	"claude-2.0":                          4,     // $8 / 1M tokens
-	"claude-2.1":                          4,     // $8 / 1M tokens
-	"claude-3-haiku-20240307":             0.125, // $0.25 / 1M tokens
-	"claude-3-5-haiku-20241022":           0.5,   // $1 / 1M tokens
-	"claude-3-sonnet-20240229":            1.5,   // $3 / 1M tokens
-	"claude-3-5-sonnet-20240620":          1.5,
-	"claude-3-5-sonnet-20241022":          1.5,
-	"claude-3-7-sonnet-20250219":          1.5,
-	"claude-3-7-sonnet-20250219-thinking": 1.5,
-	"claude-3-opus-20240229":              7.5, // $15 / 1M tokens
-	"ERNIE-4.0-8K":                        0.120 * RMB,
-	"ERNIE-3.5-8K":                        0.012 * RMB,
-	"ERNIE-3.5-8K-0205":                   0.024 * RMB,
-	"ERNIE-3.5-8K-1222":                   0.012 * RMB,
-	"ERNIE-Bot-8K":                        0.024 * RMB,
-	"ERNIE-3.5-4K-0205":                   0.012 * RMB,
-	"ERNIE-Speed-8K":                      0.004 * RMB,
-	"ERNIE-Speed-128K":                    0.004 * RMB,
-	"ERNIE-Lite-8K-0922":                  0.008 * RMB,
-	"ERNIE-Lite-8K-0308":                  0.003 * RMB,
-	"ERNIE-Tiny-8K":                       0.001 * RMB,
-	"BLOOMZ-7B":                           0.004 * RMB,
-	"Embedding-V1":                        0.002 * RMB,
-	"bge-large-zh":                        0.002 * RMB,
-	"bge-large-en":                        0.002 * RMB,
-	"tao-8k":                              0.002 * RMB,
-	"PaLM-2":                              1,
-	"gemini-1.5-pro-latest":               1.25, // $3.5 / 1M tokens
-	"gemini-1.5-flash-latest":             0.075,
-	"gemini-2.0-flash":                    0.05,
-	"gemini-2.5-pro-exp-03-25":            0.625,
-	"gemini-2.5-pro-preview-03-25":        0.625,
-	"text-embedding-004":                  0.001,
-	"chatglm_turbo":                       0.3572,     // ￥0.005 / 1k tokens
-	"chatglm_pro":                         0.7143,     // ￥0.01 / 1k tokens
-	"chatglm_std":                         0.3572,     // ￥0.005 / 1k tokens
-	"chatglm_lite":                        0.1429,     // ￥0.002 / 1k tokens
-	"glm-4":                               7.143,      // ￥0.1 / 1k tokens
-	"glm-4v":                              0.05 * RMB, // ￥0.05 / 1k tokens
-	"glm-4-alltools":                      0.1 * RMB,  // ￥0.1 / 1k tokens
-	"glm-3-turbo":                         0.3572,
-	"glm-4-plus":                          0.05 * RMB,
-	"glm-4-0520":                          0.1 * RMB,
-	"glm-4-air":                           0.001 * RMB,
-	"glm-4-airx":                          0.01 * RMB,
-	"glm-4-long":                          0.001 * RMB,
-	"glm-4-flash":                         0,
-	"glm-4v-plus":                         0.01 * RMB,
-	"qwen-turbo":                          0.8572, // ￥0.012 / 1k tokens
-	"qwen-plus":                           10,     // ￥0.14 / 1k tokens
-	"text-embedding-v1":                   0.05,   // ￥0.0007 / 1k tokens
-	"SparkDesk-v1.1":                      1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v2.1":                      1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v3.1":                      1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v3.5":                      1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v4.0":                      1.2858,
-	"360GPT_S2_V9":                        0.8572, // ¥0.012 / 1k tokens
-	"360gpt-turbo":                        0.0858, // ¥0.0012 / 1k tokens
-	"360gpt-turbo-responsibility-8k":      0.8572, // ¥0.012 / 1k tokens
-	"360gpt-pro":                          0.8572, // ¥0.012 / 1k tokens
-	"360gpt2-pro":                         0.8572, // ¥0.012 / 1k tokens
-	"embedding-bert-512-v1":               0.0715, // ¥0.001 / 1k tokens
-	"embedding_s1_v1":                     0.0715, // ¥0.001 / 1k tokens
-	"semantic_similarity_s1_v1":           0.0715, // ¥0.001 / 1k tokens
-	"hunyuan":                             7.143,  // ¥0.1 / 1k tokens  // https://cloud.tencent.com/document/product/1729/97731#e0e6be58-60c8-469f-bdeb-6c264ce3b4d0
+	"text-davinci-edit-001":                     10,
+	"code-davinci-edit-001":                     10,
+	"whisper-1":                                 15,  // $0.006 / minute -> $0.006 / 150 words -> $0.006 / 200 tokens -> $0.03 / 1k tokens
+	"tts-1":                                     7.5, // 1k characters -> $0.015
+	"tts-1-1106":                                7.5, // 1k characters -> $0.015
+	"tts-1-hd":                                  15,  // 1k characters -> $0.03
+	"tts-1-hd-1106":                             15,  // 1k characters -> $0.03
+	"davinci":                                   10,
+	"curie":                                     10,
+	"babbage":                                   10,
+	"ada":                                       10,
+	"text-embedding-3-small":                    0.01,
+	"text-embedding-3-large":                    0.065,
+	"text-embedding-ada-002":                    0.05,
+	"text-search-ada-doc-001":                   10,
+	"text-moderation-stable":                    0.1,
+	"text-moderation-latest":                    0.1,
+	"claude-instant-1":                          0.4,   // $0.8 / 1M tokens
+	"claude-2.0":                                4,     // $8 / 1M tokens
+	"claude-2.1":                                4,     // $8 / 1M tokens
+	"claude-3-haiku-20240307":                   0.125, // $0.25 / 1M tokens
+	"claude-3-5-haiku-20241022":                 0.5,   // $1 / 1M tokens
+	"claude-3-sonnet-20240229":                  1.5,   // $3 / 1M tokens
+	"claude-3-5-sonnet-20240620":                1.5,
+	"claude-3-5-sonnet-20241022":                1.5,
+	"claude-3-7-sonnet-20250219":                1.5,
+	"claude-3-7-sonnet-20250219-thinking":       1.5,
+	"claude-sonnet-4-20250514":                  1.5,
+	"claude-3-opus-20240229":                    7.5, // $15 / 1M tokens
+	"claude-opus-4-20250514":                    7.5,
+	"ERNIE-4.0-8K":                              0.120 * RMB,
+	"ERNIE-3.5-8K":                              0.012 * RMB,
+	"ERNIE-3.5-8K-0205":                         0.024 * RMB,
+	"ERNIE-3.5-8K-1222":                         0.012 * RMB,
+	"ERNIE-Bot-8K":                              0.024 * RMB,
+	"ERNIE-3.5-4K-0205":                         0.012 * RMB,
+	"ERNIE-Speed-8K":                            0.004 * RMB,
+	"ERNIE-Speed-128K":                          0.004 * RMB,
+	"ERNIE-Lite-8K-0922":                        0.008 * RMB,
+	"ERNIE-Lite-8K-0308":                        0.003 * RMB,
+	"ERNIE-Tiny-8K":                             0.001 * RMB,
+	"BLOOMZ-7B":                                 0.004 * RMB,
+	"Embedding-V1":                              0.002 * RMB,
+	"bge-large-zh":                              0.002 * RMB,
+	"bge-large-en":                              0.002 * RMB,
+	"tao-8k":                                    0.002 * RMB,
+	"PaLM-2":                                    1,
+	"gemini-1.5-pro-latest":                     1.25, // $3.5 / 1M tokens
+	"gemini-1.5-flash-latest":                   0.075,
+	"gemini-2.0-flash":                          0.05,
+	"gemini-2.5-pro-exp-03-25":                  0.625,
+	"gemini-2.5-pro-preview-03-25":              0.625,
+	"gemini-2.5-flash-preview-04-17":            0.075,
+	"gemini-2.5-flash-preview-04-17-thinking":   0.075,
+	"gemini-2.5-flash-preview-04-17-nothinking": 0.075,
+	"text-embedding-004":                        0.001,
+	"chatglm_turbo":                             0.3572,     // ￥0.005 / 1k tokens
+	"chatglm_pro":                               0.7143,     // ￥0.01 / 1k tokens
+	"chatglm_std":                               0.3572,     // ￥0.005 / 1k tokens
+	"chatglm_lite":                              0.1429,     // ￥0.002 / 1k tokens
+	"glm-4":                                     7.143,      // ￥0.1 / 1k tokens
+	"glm-4v":                                    0.05 * RMB, // ￥0.05 / 1k tokens
+	"glm-4-alltools":                            0.1 * RMB,  // ￥0.1 / 1k tokens
+	"glm-3-turbo":                               0.3572,
+	"glm-4-plus":                                0.05 * RMB,
+	"glm-4-0520":                                0.1 * RMB,
+	"glm-4-air":                                 0.001 * RMB,
+	"glm-4-airx":                                0.01 * RMB,
+	"glm-4-long":                                0.001 * RMB,
+	"glm-4-flash":                               0,
+	"glm-4v-plus":                               0.01 * RMB,
+	"qwen-turbo":                                0.8572, // ￥0.012 / 1k tokens
+	"qwen-plus":                                 10,     // ￥0.14 / 1k tokens
+	"text-embedding-v1":                         0.05,   // ￥0.0007 / 1k tokens
+	"SparkDesk-v1.1":                            1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v2.1":                            1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v3.1":                            1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v3.5":                            1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v4.0":                            1.2858,
+	"360GPT_S2_V9":                              0.8572, // ¥0.012 / 1k tokens
+	"360gpt-turbo":                              0.0858, // ¥0.0012 / 1k tokens
+	"360gpt-turbo-responsibility-8k":            0.8572, // ¥0.012 / 1k tokens
+	"360gpt-pro":                                0.8572, // ¥0.012 / 1k tokens
+	"360gpt2-pro":                               0.8572, // ¥0.012 / 1k tokens
+	"embedding-bert-512-v1":                     0.0715, // ¥0.001 / 1k tokens
+	"embedding_s1_v1":                           0.0715, // ¥0.001 / 1k tokens
+	"semantic_similarity_s1_v1":                 0.0715, // ¥0.001 / 1k tokens
+	"hunyuan":                                   7.143,  // ¥0.1 / 1k tokens  // https://cloud.tencent.com/document/product/1729/97731#e0e6be58-60c8-469f-bdeb-6c264ce3b4d0
 	// https://platform.lingyiwanwu.com/docs#-计费单元
 	// 已经按照 7.2 来换算美元价格
 	"yi-34b-chat-0205":       0.18,
@@ -252,10 +258,11 @@ var defaultCompletionRatio = map[string]float64{
 	"gpt-4-gizmo-*":  2,
 	"gpt-4o-gizmo-*": 3,
 	"gpt-4-all":      2,
+	"gpt-image-1":    8,
 }
 
-// InitModelSettings initializes all model related settings maps
-func InitModelSettings() {
+// InitRatioSettings initializes all model related settings maps
+func InitRatioSettings() {
 	// Initialize modelPriceMap
 	modelPriceMapMutex.Lock()
 	modelPriceMap = defaultModelPrice
@@ -276,7 +283,11 @@ func InitModelSettings() {
 	cacheRatioMap = defaultCacheRatio
 	cacheRatioMapMutex.Unlock()
 
-	common.SysLog("model settings initialized")
+	// initialize imageRatioMap
+	imageRatioMapMutex.Lock()
+	imageRatioMap = defaultImageRatio
+	imageRatioMapMutex.Unlock()
+
 }
 
 func GetModelPriceMap() map[string]float64 {
@@ -431,13 +442,15 @@ func getHardcodedCompletionModelRatio(name string) (float64, bool) {
 	if name == "chatgpt-4o-latest" {
 		return 3, true
 	}
-	if strings.Contains(name, "claude-instant-1") {
-		return 3, true
-	} else if strings.Contains(name, "claude-2") {
-		return 3, true
-	} else if strings.Contains(name, "claude-3") {
+
+	if strings.Contains(name, "claude-3") {
 		return 5, true
+	} else if strings.Contains(name, "claude-sonnet-4") || strings.Contains(name, "claude-opus-4") {
+		return 5, true
+	} else if strings.Contains(name, "claude-instant-1") || strings.Contains(name, "claude-2") {
+		return 3, true
 	}
+
 	if strings.HasPrefix(name, "gpt-3.5") {
 		if name == "gpt-3.5-turbo" || strings.HasSuffix(name, "0125") {
 			// https://openai.com/blog/new-embedding-models-and-api-updates
@@ -459,6 +472,12 @@ func getHardcodedCompletionModelRatio(name string) (float64, bool) {
 			return 4, true
 		} else if strings.HasPrefix(name, "gemini-2.5-pro-preview") {
 			return 8, true
+		} else if strings.HasPrefix(name, "gemini-2.5-flash-preview") {
+			if strings.HasSuffix(name, "-nothinking") {
+				return 4, false
+			} else {
+				return 3.5 / 0.6, false
+			}
 		}
 		return 4, false
 	}
@@ -502,18 +521,18 @@ func getHardcodedCompletionModelRatio(name string) (float64, bool) {
 
 func GetAudioRatio(name string) float64 {
 	if strings.Contains(name, "-realtime") {
-		if strings.HasSuffix(name, "gpt-4o-realtime-preview-2024-12-17") {
+		if strings.HasSuffix(name, "gpt-4o-realtime-preview") {
 			return 8
-		} else if strings.Contains(name, "mini") {
+		} else if strings.Contains(name, "gpt-4o-mini-realtime-preview") {
 			return 10 / 0.6
 		} else {
 			return 20
 		}
 	}
 	if strings.Contains(name, "-audio") {
-		if strings.HasSuffix(name, "gpt-4o-audio-preview-2024-12-17") {
-			return 16
-		} else if strings.Contains(name, "mini") {
+		if strings.HasPrefix(name, "gpt-4o-audio-preview") {
+			return 40 / 2.5
+		} else if strings.HasPrefix(name, "gpt-4o-mini-audio-preview") {
 			return 10 / 0.15
 		} else {
 			return 40
@@ -541,3 +560,36 @@ func ModelRatio2JSONString() string {
 	}
 	return string(jsonBytes)
 }
+
+var defaultImageRatio = map[string]float64{
+	"gpt-image-1": 2,
+}
+var imageRatioMap map[string]float64
+var imageRatioMapMutex sync.RWMutex
+
+func ImageRatio2JSONString() string {
+	imageRatioMapMutex.RLock()
+	defer imageRatioMapMutex.RUnlock()
+	jsonBytes, err := json.Marshal(imageRatioMap)
+	if err != nil {
+		common.SysError("error marshalling cache ratio: " + err.Error())
+	}
+	return string(jsonBytes)
+}
+
+func UpdateImageRatioByJSONString(jsonStr string) error {
+	imageRatioMapMutex.Lock()
+	defer imageRatioMapMutex.Unlock()
+	imageRatioMap = make(map[string]float64)
+	return json.Unmarshal([]byte(jsonStr), &imageRatioMap)
+}
+
+func GetImageRatio(name string) (float64, bool) {
+	imageRatioMapMutex.RLock()
+	defer imageRatioMapMutex.RUnlock()
+	ratio, ok := imageRatioMap[name]
+	if !ok {
+		return 1, false // Default to 1 if not found
+	}
+	return ratio, true
+}

setting/operation_setting/tools.go

@@ -0,0 +1,57 @@
+package operation_setting
+
+import "strings"
+
+const (
+	// Web search
+	WebSearchHighTierModelPriceLow    = 30.00
+	WebSearchHighTierModelPriceMedium = 35.00
+	WebSearchHighTierModelPriceHigh   = 50.00
+	WebSearchPriceLow                 = 25.00
+	WebSearchPriceMedium              = 27.50
+	WebSearchPriceHigh                = 30.00
+	// File search
+	FileSearchPrice = 2.5
+)
+
+func GetWebSearchPricePerThousand(modelName string, contextSize string) float64 {
+	// 确定模型类型
+	// https://platform.openai.com/docs/pricing Web search 价格按模型类型和 search context size 收费
+	// gpt-4.1, gpt-4o, or gpt-4o-search-preview 更贵，gpt-4.1-mini, gpt-4o-mini, gpt-4o-mini-search-preview 更便宜
+	isHighTierModel := (strings.HasPrefix(modelName, "gpt-4.1") || strings.HasPrefix(modelName, "gpt-4o")) &&
+		!strings.Contains(modelName, "mini")
+	// 确定 search context size 对应的价格
+	var priceWebSearchPerThousandCalls float64
+	switch contextSize {
+	case "low":
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceLow
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceLow
+		}
+	case "medium":
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceMedium
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceMedium
+		}
+	case "high":
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceHigh
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceHigh
+		}
+	default:
+		// search context size 默认为 medium
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceMedium
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceMedium
+		}
+	}
+	return priceWebSearchPerThousandCalls
+}
+
+func GetFileSearchPricePerThousand() float64 {
+	return FileSearchPrice
+}

setting/rate_limit.go

@@ -1,6 +1,64 @@
 package setting
 
+import (
+	"encoding/json"
+	"fmt"
+	"one-api/common"
+	"sync"
+)
+
 var ModelRequestRateLimitEnabled = false
 var ModelRequestRateLimitDurationMinutes = 1
 var ModelRequestRateLimitCount = 0
 var ModelRequestRateLimitSuccessCount = 1000
+var ModelRequestRateLimitGroup = map[string][2]int{}
+var ModelRequestRateLimitMutex sync.RWMutex
+
+func ModelRequestRateLimitGroup2JSONString() string {
+	ModelRequestRateLimitMutex.RLock()
+	defer ModelRequestRateLimitMutex.RUnlock()
+
+	jsonBytes, err := json.Marshal(ModelRequestRateLimitGroup)
+	if err != nil {
+		common.SysError("error marshalling model ratio: " + err.Error())
+	}
+	return string(jsonBytes)
+}
+
+func UpdateModelRequestRateLimitGroupByJSONString(jsonStr string) error {
+	ModelRequestRateLimitMutex.RLock()
+	defer ModelRequestRateLimitMutex.RUnlock()
+
+	ModelRequestRateLimitGroup = make(map[string][2]int)
+	return json.Unmarshal([]byte(jsonStr), &ModelRequestRateLimitGroup)
+}
+
+func GetGroupRateLimit(group string) (totalCount, successCount int, found bool) {
+	ModelRequestRateLimitMutex.RLock()
+	defer ModelRequestRateLimitMutex.RUnlock()
+
+	if ModelRequestRateLimitGroup == nil {
+		return 0, 0, false
+	}
+
+	limits, found := ModelRequestRateLimitGroup[group]
+	if !found {
+		return 0, 0, false
+	}
+	return limits[0], limits[1], true
+}
+
+func CheckModelRequestRateLimitGroup(jsonStr string) error {
+	checkModelRequestRateLimitGroup := make(map[string][2]int)
+	err := json.Unmarshal([]byte(jsonStr), &checkModelRequestRateLimitGroup)
+	if err != nil {
+		return err
+	}
+	for group, limits := range checkModelRequestRateLimitGroup {
+		if limits[0] < 0 || limits[1] < 1 {
+			return fmt.Errorf("group %s has negative rate limit values: [%d, %d]", group, limits[0], limits[1])
+		}
+	}
+
+	return nil
+}

setting/system_setting.go

@@ -3,6 +3,7 @@ package setting
 var ServerAddress = "http://localhost:3000"
 var WorkerUrl = ""
 var WorkerValidKey = ""
+var WorkerAllowHttpImageRequestEnabled = false
 
 func EnableWorker() bool {
 	return WorkerUrl != ""

web/.prettierrc.mjs

@@ -1 +1 @@
-module.exports = require("@so1ve/prettier-config");
+module.exports = require('@so1ve/prettier-config');