NextChat
NextChat copied to clipboard
ChatGPTNextWeb
[Bug] 自定义模型分组无法使用，报错 you are not allowed to use xxx model
📦 Deployment Method
Docker
📌 Version
v2.16.0
💻 Operating System
Ubuntu
📌 System Version
22.04
🌐 Browser
Chrome
📌 Browser Version
latest
🐛 Bug Description
使用背景
我目前在使用 v2.15.3，docker 部署方式
各个渠道的key都使用 one-api 管理，然后在启动 next-chat 时设置
-e CUSTOM_MODELS="-all,+gpt-3.5-turbo@AZ-OAI,AZ/DeepSeek-R1@AZ-AI,gpt-4o@openai,claude-3-5-sonnet@BP-Anthropic,Qwen/Qwen2.5-7B-Instruct@硅基流动,THUDM/glm-4-9b-chat@硅基流动,glm-4-flash@智谱"
这个 model_name@group_name 的方式在UI展示上很好用
错误描述
用同样的 CUSTOM_MODELS 启动，最新的版本里UI显示没问题，但使用自定义模型时会报错：
{
"error": true,
"message": "you are not allowed to use xxx model"
}
只有后面 @openai,@OpenAI 或者不加@xxx 的自定义模型可以用
预期
希望能够继续使用 model_name@group_name 的方式来展示和使用
📷 Recurrence Steps
No response
🚦 Expected Behavior
No response
📝 Additional Information
No response
Bot detected the issue body's language is not English, translate it automatically.
I'm the same problem, and I can use it after removing @
modelname@provider格式中后面的provider是指定服务商的，不是用来随便指定分组打标签的。
目前只有内置支持的服务商可以选择，比如OpenAI，Anthropic，Deepseek等等，你使用one-api中转，按照OpenAI的格式转发，那自定义模型只能选择@OpenAI，或者按照默认值不写。
Bot detected the issue body's language is not English, translate it automatically.
The provider in the modelname@provider format is specified by the service provider, not used to specify groups and label them casually.
Currently, only service providers with built-in support can choose, such as OpenAI, Anthropic, Deepseek, etc. If you use one-api relay and forward according to the OpenAI format, then the custom model can only choose @OpenAI, or do not write according to the default value.
Bot detected the issue body's language is not English, translate it automatically.
Since @openai is considered a channel method selection, can another method be provided as model grouping identification?
Bot detected the issue body's language is not English, translate it automatically.
It was possible, but the pr that fixed this bug was put on hold
Bot detected the issue body's language is not English, translate it automatically.
If you need to display, you can use displayName
modelname@provider格式中后面的provider是指定服务商的，不是用来随便指定分组打标签的。
目前只有内置支持的服务商可以选择，比如OpenAI，Anthropic，Deepseek等等，你使用one-api中转，按照OpenAI的格式转发，那自定义模型只能选择
@OpenAI，或者按照默认值不写。
也就是说之前版本能够使用这个方式打标签才是bug？😂 也许是feature吧
Bot detected the issue body's language is not English, translate it automatically.
The provider in the modelname@provider format is specified by the service provider, not used to specify groups and label them casually.
Currently, only service providers with built-in support can choose, such as OpenAI, Anthropic, Deepseek, etc. If you use one-api to forward according to the OpenAI format, then the custom model can only choose
@OpenAI, or according to the default value, Write.
In other words, the previous version was able to use this method to label it was a bug? 😂
本来是可以的，但是修复这个 bug 的 pr 被搁置了
#6000 看到了，还是希望能支持自定义打标签功能，比较方便。displayName只是修改了模型名，后面括号里的provider没变，没起到分类的目的
Bot detected the issue body's language is not English, translate it automatically.
It was possible, but the pr that fixed this bug was put on hold
#6000 I saw it, but I still hope it can support the custom tagging function, which is more convenient. displayName only modified the model name, the provider in the brackets afterwards did not change, and did not serve the purpose of classification.
这个还是不行吗？
配置：
services:
chatgpt-next-web:
image: yidadaa/chatgpt-next-web:v2.16.0
container_name: chatgpt-next-web
ports:
- "3000:3000"
restart: always
environment:
OPENAI_API_KEY: sk-xxx-xxx
PROXY_URL: xxx
DEFAULT_MODEL: gpt-4o
BYTEDANCE_URL: https://ark.cn-beijing.volces.com/api/v3
BYTEDANCE_API_KEY: xxx-xxx
CUSTOM_MODELS: +deepseek-v3-241226@bytedance=deepseek-v3-241226
无法用，出现异常
Bot detected the issue body's language is not English, translate it automatically.
Is this still not possible?
Configuration:
services:
chatgpt-next-web:
image: yidadaa/chatgpt-next-web:v2.16.0
container_name: chatgpt-next-web
Ports:
- "3000:3000"
restart: always
environment:
OPENAI_API_KEY: sk-xxx-xxx
PROXY_URL: xxx
DEFAULT_MODEL: gpt-4o
BYTEDANCE_URL: https://ark.cn-beijing.volces.com/api/v3
BYTEDANCE_API_KEY: xxx-xxx
CUSTOM_MODELS: +deepseek-v3-241226@bytedance=deepseek-v3-241226
Unable to use, an exception occurs