[Bug] Invalid URL (POST /v1/chat/completions)

[johnfelipe]

latest commit show me this issue:

latest one of that is OK:

[Yidadaa]

have you set Endpoint?

[Issues-translate-bot]

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

---

have you set Endpoint?

[johnfelipe]

What need to put in that field? El mar, 27 jun 2023, 9:49 p. m., Issues-translate-bot < ***@***.***> escribió:

…

Bot detected the issue body's language is not English, translate it automatically. ------------------------------ have you set Endpoint? [image: image] <https://user-images.githubusercontent.com/16968934/249331736-d506ffe7-ff98-4df8-ae62-c0d9bb92ded9.png> — Reply to this email directly, view it on GitHub <#2175 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADIWFCEOUCFYPG47IJMYFDXNOLTRANCNFSM6AAAAAAZWHFZNQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[Yidadaa]

If you have direct access to api.openai.com, please set it as https://api.openai.com.
If you do not have direct access to api.openai.com, please keep the option as /api/openai. This will utilize Vercel to proxy all your requests.

[Yidadaa]

Later, I will add a default logic to the code, so if the user forgets to set this option, the default endpoint will be automatically used.

[QAQQL]

当我使用 https://github.com/THUDM/ChatGLM2-6B 时,官方有推荐使用该项目
感谢 [@hiyouga](https://github.com/THUDM/ChatGLM2-6B/blob/main) 实现了 OpenAI 格式的流式 API 部署，可以作为任意基于 ChatGPT 的应用的后端，比如 [ChatGPT-Next-Web](https://github.com/Yidadaa/ChatGPT-Next-Web)。可以通过运行仓库中的[openai_api.py](https://github.com/THUDM/ChatGLM2-6B/blob/main/openai_api.py) 进行部署：

我进行了如下配置

发现下拉模型列表并不能触发/v1/models接口,且接口访问全为跨域

#1179

[QAQQL]

下面的模型无法选择

[Issues-translate-bot]

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

---

The following models cannot be selected

[Yidadaa]

@QAQQL 兼容接口无需选中 chatglm 模型，使用默认的 gpt-3.5-turbo 即可。

[Issues-translate-bot]

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

---

@QAQQL compatible interface does not need to select the chatglm model, just use the default gpt-3.5-turbo.

[QAQQL]

但是他接口访问失败

他会先调用option

[Issues-translate-bot]

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

---

But his interface access failed

[Yidadaa]

@QAQQL 这是浏览器的默认策略，你可以给他们提个 PR 来解决这个问题，只需要允许所有的 OPTIONS 请求，并返回 200 即可。

[Issues-translate-bot]

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

---

@QAQQL This is the browser's default policy, you can file a PR with them to fix this, just allow all OPTIONS requests and return 200.

[Yidadaa]

@QAQQL 你可以参考 fastapi 的 api 文档，为它添加 cors 支持：https://fastapi.tiangolo.com/tutorial/cors/#steps

只需要在：

https://github.com/THUDM/ChatGLM2-6B/blob/ba60190296e3451c0e7e6a0f26c60a8ba7720dfe/openai_api.py#L26

处，加几行代码：

from fastapi.middleware.cors import CORSMiddleware

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

[Yidadaa]

等待此 PR 合入：THUDM/ChatGLM2-6B#102

[QAQQL]

好的,谢谢您

[Issues-translate-bot]

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

---

Ok, thank you

[KAIYI-HSU]

@Yidadaa
您好:

我也是透過ChatGLM2-6B 的openai_api.py進行部署後端。
我只修改文件中的model路徑跟改對外端口如下:

我有確定成功開啟端口並且能成功從 <我的固定IP>:19745/docs造訪說明書，也能用POST測試成功，GLM2可以生成回應。

請問我使用您的前端網頁: https://chatgpt1.nextweb.fun/

Endpoint填寫我的<我的固定IP>:19745 or https://<我的固定IP>:19745
對話時會出現以下錯誤:
{
"error": true,
"message": "Failed to fetch"
}

=====
希望能尋求幫助，
為何我後端部署好，任何前端網頁都無法跟 我的<我的固定IP>:19745 or https://<我的固定IP>:19745
進行互動?

[Issues-translate-bot]

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

---

@Yidadaa
Hello:

I also deploy the backend through the openai_api.py of ChatGLM2-6B.
I only modify the model path in the file and change the external port as follows:

I have confirmed that the port is successfully opened and I can successfully access the manual from :19745/docs, and I can also use POST to test successfully, and GLM2 can generate a response.

Please let me use your front-end webpage: https://chatgpt1.nextweb.fun/

Endpoint fill in my :19745 or https://:19745
The following error occurs while talking:
{
"error": true,
"message": "Failed to fetch"
}

=====
Hoping for help,
Why can't any front-end webpage follow my backend deployment? My :19745 or https://:19745
to interact?