[Bug] 模型通过什么什么服务商来访问应该可以自定义，而不是固定

[Ubivision]

📦 部署方式

官方安装包

📌 软件版本

2.15.1

💻 系统环境

Windows, macOS

📌 系统版本

Sonama 14.5

🌐 浏览器

Chrome, Safari

📌 浏览器版本

safari17.5

🐛 问题描述

当自定义模型名称和已存在模型名称重复时，会强制使用模型列表中对应模型的 api 格式。比如我有自己的 oneapi 可以调用 cladue，但会被强制使用anthropic 的 api，导致 cladue 所有模型我都无法使用。

📷 复现步骤

No response

🚦 期望结果

模型可以自定义选择使用不同的模型服务商

📝 补充信息

No response

[Issues-translate-bot]

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

---

Title: [Bug] The service provider through which the model is accessed should be customizable rather than fixed.

📦 Deployment method

Official installation package

📌 Software version

2.15.1

💻 System environment

Windows, macOS

📌 System version

Sonama 14.5

🌐 Browser

Chrome, Safari

📌 Browser version

safari17.5

🐛 Problem description

When the name of a custom model overlaps with an existing model name, the api format of the corresponding model in the model list will be forced to be used. For example, I have my own oneapi that can call cladue, but it will be forced to use anthropopic's api, resulting in that I cannot use all models of cladue.

📷 Steps to reproduce

No response

🚦 Expected results

Models can be customized and choose to use different model service providers

📝 Supplementary information

No response

[Kosette]

如果想要通过 OpenAI 兼容的api调用模型，自定义模型名称应该为 <modelName>@OpenAI，比如claude-3-5-sonnet@OpenAI，同理如果调用claude兼容的api接口，则使用 <modelName>@Anthropic。详情参考置顶issue #5001

在使用对应的自定义模型前，需设置好相关的API_URL。

[Issues-translate-bot]

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

---

If you want to call the model through the OpenAI compatible API, the custom model name should be <modelName>@OpenAI, such as claude-3-5-sonnet@OpenAI. Similarly, if you call the claude compatible API interface, use <modelName>@Anthropic.

Before using the corresponding custom model, the relevant API_URL needs to be set.

[skymkmk]

请仔细阅读 #5001

[Issues-translate-bot]

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

---

Please read #5001 carefully