[Feature Request]: 能否加个选项为每个对话选择是使用自定义接口的url和key还是选择默认环境变量设置的url和key

[czh020110]

Problem Description

能否加个选项为每个对话选择是使用自定义接口的url和key还是选择默认环境变量设置的url和key，这样中转接口可以换两个不同的地址，很方便

Solution Description

/

Alternatives Considered

。

Additional Context

。

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

Title: [Feature Request]: Can you add an option for each conversation to choose whether to use the url and key of the custom interface or the url and key set by the default environment variables?

Problem Description

Can you add an option for each conversation to choose whether to use the url and key of the custom interface or the url and key set by the default environment variables, so that the transfer interface can be changed to two different addresses, which is very convenient

Solution Description

/

Alternatives Considered

.

Additional Context

.

[lloydzhou]

#4930 (comment)

或许这里有一个相对更全面一些的解决方案：

1. 如果你使用gpt-4，并且有两家供应商。
2. 就可以通过CUSTOM_PROVIDER定义两个自定义的供应商（暂时还未实现）

类型都是走ServiceProvider.OpenAI
只是配置的endpoint和apikey不一样（这代表发送到next server的地址会不一样，这样在next server中即可使用不用的apikey转发实际的请求）

3. 然后结合CUSTOM_MODELS，配置两个模型选项到模型切换列表

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

#4930 (comment)

Perhaps here is a relatively more comprehensive solution:

1. If you use gpt-4 and have two vendors.
2. You can define two custom suppliers through CUSTOM_PROVIDER (not yet implemented)

The types are all ServiceProvider.OpenAI
Only the configured endpoint and apikey are different (this means that the address sent to the next server will be different, so that the different apikey can be used in the next server to forward the actual request)

3. Then combine CUSTOM_MODELS to configure the two model options to the model switching list

[czh020110]

#4930 (comment)

或许这里有一个相对更全面一些的解决方案：

1. 如果你使用gpt-4，并且有两家供应商。
2. 就可以通过CUSTOM_PROVIDER定义两个自定义的供应商（暂时还未实现）

类型都是走ServiceProvider.OpenAI
只是配置的endpoint和apikey不一样（这代表发送到next server的地址会不一样，这样在next server中即可使用不用的apikey转发实际的请求）

3. 然后结合CUSTOM_MODELS，配置两个模型选项到模型切换列表

确实，这种方法更加灵活！自定义性更强！

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically.

---

#4930 (comment)

Perhaps here is a relatively more comprehensive solution:

1. If you use gpt-4 and have two vendors.
2. You can define two custom suppliers through CUSTOM_PROVIDER (not yet implemented)

The types are all ServiceProvider.OpenAI
Only the configured endpoint and apikey are different (this means that the address sent to the next server will be different, so that the different apikey can be used in the next server to forward the actual request)

3. Then combine CUSTOM_MODELS to configure the two model options to the model switching list

Indeed, this approach is more flexible! More customizable!