[Bug] Azure configure

[kagoesy]

📦 部署方式

Docker

📌 软件版本

2.13.1

💻 系统环境

Windows

📌 系统版本

all

🌐 浏览器

Chrome, Edge

📌 浏览器版本

unknown

🐛 问题描述

在最近的更新后，使用custom endpoint配置Azure OpenAI时，model会被自动填充成endpoint中deployments的名字且无法修改，但是在部署Azure OpenAI deployment时名字可以是多样的，导致当没有使用标准的model名字作为deployment名字时发送消息会一直报404:

📷 复现步骤

1. 部署deployment名字为gpt4oChris
   
2. 将endpoint及其他信息填入custom endpoint, 其中endpoint内容为“https://xxxxx.openai.azure.com/openai/deployments/gpt4oChris“ 且model选为gpt-4o(Azure):
   
3. 发送消息得到404
   
4. 通过浏览器开发者模式可以看到请求发送的地址变成了"https://xxxxx.openai.azure.com/openai/deployments/gpt4oChris/deployments/gpt-4o"，多增加了有model自动写入的错误deployment名字
   

🚦 期望结果

1. 如果按照提示的格式带有deployments id填写endpoint，则不应该在之后自动添加带有model名字的deployment id

或者

2. 如果不希望在endpoint时加入deployment id且要自动生成的话，可以使用custom model来填写deployment id，但是希望添加说明，以及根据测试，当前版本添加custom model时仍旧添加的是标准的模型名字为deployment id

📝 补充信息

No response

[nextchat-manager]

Please follow the issue template to update description of your issue.

[Issues-translate-bot]

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

---

Title: [Bug]

📦 Deployment method

Docker

📌 Software version

2.13.1

💻 System environment

Windows

📌 System version

all

🌐 Browser

Chrome, Edge

📌 Browser version

unknown

🐛 Problem description

After the recent update, when configuring Azure OpenAI using a custom endpoint, the model will be automatically filled in with the name of the deployments in the endpoint and cannot be modified. However, when deploying Azure OpenAI deployment, the name can be diverse, resulting in when the standard model name is not used. When sending a message as a deployment name, 404 will always be reported:

📷 Steps to reproduce

1. Deployment deployment name is gpt4oChris
   
2. Fill in the endpoint and other information into the custom endpoint, where the endpoint content is "https://xxxxx.openai.azure.com/openai/deployments/gpt4oChris" and the model is selected as gpt-4o (Azure):
   
3. Send message and get 404
   
4. Through the browser developer mode, you can see that the address sent by the request has changed to "https://xxxxx.openai.azure.com/openai/deployments/gpt4oChris/deployments/gpt-4o", and a model has been added to automatically Error deployment name written
   

🚦 Expected results

1. If you fill in the endpoint with the deployment id according to the prompt format, you should not automatically add the deployment id with the model name later.

or

2. If you do not want to add the deployment id to the endpoint and want to automatically generate it, you can use a custom model to fill in the deployment id, but you want to add a description, and according to the test, when adding the custom model in the current version, the standard model name is still added. deployment id

📝 Supplementary information

No response

[lloydzhou]

#5001

[kagoesy]

#5001

懂了，原来是语法拓展了，那是不是endpoint的example提示可以修改一下以防产生歧义：

[lloydzhou]

拓展语法的原因是为了使用Azure的时候，能使用多个不同的模型（之前的旧版，实际上只能使用单个版本，不管怎么切换其实都不起作用）

关于这里的example确实需要改

[Issues-translate-bot]

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

---

The reason for expanding the syntax is to be able to use multiple different models when using Azure (the previous old version could actually only use a single version, and it would not work no matter how you switch)

The example here really needs to be changed

[ErYiii]

#5001

懂了，原来是语法拓展了，那是不是endpoint的example提示可以修改一下以防产生歧义：

你好，请问新的API格式应该是怎样的呢，我也是用的AzureOpenAI，更新2.14后用不了了。。

[Issues-translate-bot]

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

---

#5001

Got it, it turns out that the syntax has been expanded. Can the example prompt of the endpoint be modified to prevent ambiguity: 

Hello, what should the new API format be? I also use AzureOpenAI, but it can no longer be used after updating to 2.14. .

[kagoesy]

#5001

懂了，原来是语法拓展了，那是不是endpoint的example提示可以修改一下以防产生歧义：

你好，请问新的API格式应该是怎样的呢，我也是用的AzureOpenAI，更新2.14后用不了了。。

我理解的就是endpoint就只保留到openai就可以，不要加deployment ID，然后custom model要写成 model@azure=deployment ID的形式

[Issues-translate-bot]

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

---

#5001

Got it, it turns out that the syntax has been expanded. Can the example prompt of the endpoint be modified to prevent ambiguity: 

Hello, what should the new API format be? I also use AzureOpenAI, but it can no longer be used after updating to 2.14. .

What I understand is that the endpoint can only be kept in OpenAI, do not add deployment ID, and then the custom model should be written in the form of model@azure=

[ErYiii]

#5001

懂了，原来是语法拓展了，那是不是endpoint的example提示可以修改一下以防产生歧义：

你好，请问新的API格式应该是怎样的呢，我也是用的AzureOpenAI，更新2.14后用不了了。。

我理解的就是endpoint就只保留到openai就可以，不要加deployment ID，然后custom model要写成 model@azure=deployment ID的形式

是的，中午琢磨半天，已经解决了。谢谢。

[Issues-translate-bot]

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

---

#5001

I understand, it turns out that the syntax has been expanded. Is it the endpoint’s example prompt that can be modified to prevent ambiguity: 

Hello, what should the new API format be? I also use AzureOpenAI, but it can no longer be used after updating to 2.14. .

What I understand is that the endpoint can only be kept in openai, do not add deployment ID, and then the custom model should be written in the form of model@azure=deployment ID

Yes, after thinking about it for a long time at noon, I have solved it. Thanks.