Allow anonymous access to my Azure OpenAI chat bot

题意:允许匿名访问我的 Azure OpenAI 聊天机器人

问题背景:

I have an Azure OpenAI chat bot using my own data (I configured an OpenAI resource and chose Deploy as Web App) . Members of my domain can access it by logging in. Now I want it to be accessible anonymously or for other domains I define - whatever is easier to configure.

我有一个使用我自己数据的 Azure OpenAI 聊天机器人(我配置了一个 OpenAI 资源并选择了以 Web 应用的形式部署)。我的域内成员可以通过登录访问它。现在我希望它可以匿名访问,或者对我定义的其他域开放------以更容易配置的方式为准。

In the Azure Web App Authentication settings I could do two things:

在 Azure Web 应用的身份验证设置中,我可以做两件事:

  • Disable Authentication 禁用身份验证
  • Allow unauthenticated access 允许未认证的访问

See this screenshot: 看下面的截图

When I disable authentication or when I allow unauthenticated access I get in both cases the same error message when opening the chat bot:

当我禁用身份验证或允许未认证的访问时,在打开聊天机器人时,两种情况下都会出现相同的错误消息:

Authentication Not Configured

I also tried to allow not only my single domain but multiple account types of the Identity Provider. But when I login with a different domain I still the an error messge:

我还尝试允许不仅是我的单一域名,还包括多个身份提供者的帐户类型。但是,当我使用不同的域名登录时,我仍然看到错误消息:

AADSTS50020: User account 'myuser@otherdomain' from identity provider 'https://sts.windows.net/....../' does not exist in tenant 'MyDomain' and cannot access the application 'MyChatBot') in that tenant.

How to do it properly? 该如何正确配置?

问题解决:

Context

You have to understand that the "Deploy to Web App" button in Azure OpenAI Studio is only an accelerator for the "all-in-one" deployment of a project which is in fact open-sourced by Microsoft.

你需要了解,Azure OpenAI Studio 中的"部署到 Web 应用"按钮只是一个用于项目"一体化"部署的加速器,而该项目实际上是由微软开源的。

It's this one: 就是这个: GitHub - microsoft/sample-app-aoai-chatGPT: Sample code for a simple web chat experience through Azure OpenAI, including Azure OpenAI On Your Data.

So the Authentication Not Configured message you are facing is implemented inside this project, here:

所以你遇到的 "Authentication Not Configured" 错误消息是实现于该项目内部的,位置在这里:sample-app-aoai-chatGPT/frontend/src/pages/chat/Chat.tsx at af9d06b4ba153523fed366d3f8b1b19d1fc5f5e8 · microsoft/sample-app-aoai-chatGPT · GitHub

Possible solution 可能的解决方案

The README of the project explains that you can remove it if you want, here

该项目的 README 中解释了如果你想移除它,可以查看这里。

复制代码
To remove this restriction, you can add AUTH_ENABLED=False to the environment variables. This will disable authentication and allow anyone to access the chat functionality of your app. This is not recommended for production apps.

As said, "This is not recommended for production apps.".

正如所说,"这不建议用于生产环境的应用程序。"

Especially as the frontend is calling itself for the conversation (POST to /conversation) so if you remove the authentication, I guess that someone could use this endpoint directly to consume your OpenAI resource which is behind (ok, they can't modify your system message etc. but it's still an issue)

特别是因为前端是通过自身调用进行对话的(POST 到 /conversation),所以如果你移除了身份验证,我猜测有人可能会直接使用这个端点来消耗你后台的 OpenAI 资源(虽然他们不能修改你的系统消息等,但这仍然是一个问题)。

相关推荐
uccs2 小时前
流式响应的三次进化:EventSource → ReadableStream → TransformStream
openai·ai编程·claude
宅小年6 小时前
Codex Skills 怎么选?我常用的几个推荐给你
openai
机器之心7 小时前
近80年后,埃尔德什经典「拉姆齐数下界」,被三位中国学者首次指数级改进
人工智能·openai
机器之心7 小时前
Nvidia都在点赞的LoopWM世界模型,竟然来自一家中国初创FaceMind?
人工智能·openai
武子康15 小时前
调查研究-202 SGLang 深度解析:为什么大模型推理框架不只是“把模型跑起来“
人工智能·openai·agent
怕浪猫16 小时前
第6章 多智能体协作:从单兵作战到群体智能
aigc·openai·ai编程
doiito17 小时前
【Agent Harness】Gliding Horse 的 L2 作战地图:让多 Agent 协作从“摸黑”变成“透明”
ai·rust·架构设计·系统设计·ai agent
xiezhr17 小时前
逛GitHub发现一款免费带有AI功能的数据库管理工具DBX
ai·开源软件·自然语言·数据库管理工具
山间小僧17 小时前
「AI学习笔记」Harness Engineering: 从 Claude Code 源码看 Harness Engineering 最佳实践
aigc·openai·ai编程
用户47949283569152 天前
翻完 lark-cli 的 17 万行 Go 代码,我学到了什么
后端·openai