[Bug] 把glm-4-flash设置为默认模型，但是依旧访问openapi的接口

[chunxingque]

📦 部署方式

官方安装包

📌 软件版本

v2.15.8

💻 系统环境

Other Linux

📌 系统版本

centos7.9

🌐 浏览器

Chrome

📌 浏览器版本

131.0.6778.264

🐛 问题描述

1.我使用docker部署，我想默认使用glm-4-flash模型，部署命令如下

docker run --rm -p 3000:3000 --name chatgpt-next-web \
   -e OPENAI_API_KEY=************ \
   -e CHATGLM_API_KEY=*************** \
   -e CHATGLM_URL=https://open.bigmodel.cn \
   -e DEFAULT_MODEL=glm-4-flash \
   -e CUSTOM_MODELS=-all,glm-4-flash \
   -e DEFAULT_INPUT_TEMPLATE=0.95 \
   -e DISABLE_GPT4=1 \
   -v /etc/localtime:/etc/localtime:ro  \
yidadaa/chatgpt-next-web:v2.15.8

2.我打开网页后，输入文本时，前台显示连接错误

3.我看了运行日志，显示调用的是openapi的api地址，而不是glm-4-flash模型的api地址

4.我看了当前会话，显示的模型是glm-4-flash，但是依旧访问不了

我怀疑是浏览器的缓存问题，但是我打开无痕窗口重新访问，接口依旧访问到openapi的api地址，而不是glm-4-flash模型的api地址

📷 复现步骤

无

🚦 期望结果

设置glm-4-flash为默认模型，打开的聊天，使用的是glm-4-flash模型，而不是openapi模型，并且可以正常使用

📝 补充信息

无

[Issues-translate-bot]

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

---

Title: [Bug] Set glm-4-flash as the default model, but still access the openapi interface

📦 Deployment method

Official installation package

📌 Software version

v2.15.8

💻 System environment

Other Linux

📌 System version

centos7.9

🌐 Browser

Chrome

📌 Browser version

131.0.6778.264

🐛 Problem description

1. I use docker to deploy. I want to use the glm-4-flash model by default. The deployment command is as follows

docker run --rm -p 3000:3000 --name chatgpt-next-web \
   -e OPENAI_API_KEY=************ \
   -e CHATGLM_API_KEY=**************** \
   -e CHATGLM_URL=https://open.bigmodel.cn \
   -e DEFAULT_MODEL=glm-4-flash \
   -e CUSTOM_MODELS=-all,glm-4-flash \
   -e DEFAULT_INPUT_TEMPLATE=0.95 \
   -e DISABLE_GPT4=1 \
   -v /etc/localtime:/etc/localtime:ro \
yidadaa/chatgpt-next-web:v2.15.8

2. After I open the web page and enter text, a connection error is displayed in the foreground.
   

3. I looked at the running log and it showed that the api address of openapi was called, not the api address of the glm-4-flash model.
   

4. I checked the current session and the displayed model is glm-4-flash, but I still cannot access it.
   

I suspected that it was a browser cache problem, but when I opened the incognito window and visited again, the interface still accessed the api address of openapi, not the api address of the glm-4-flash model.

📷 Steps to reproduce

none

🚦 Expected results

Set glm-4-flash as the default model. The opened chat uses the glm-4-flash model instead of the openapi model and can be used normally.

📝 Supplementary information

None

[TianqiZhang]

我这里有类似情况，区别是我没有设置DEFAULT_MODEL，但是在config里指定了gpt-4o@Azure。当开始一个新的对话的时候，仍然默认去访问openai的接口。模型名称是空的。每次都需要手动选择一次：

[Issues-translate-bot]

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

---

I have a similar situation here, the difference is that I did not set DEFAULT_MODEL, but specified gpt-4o@Azure in the config. When starting a new conversation, the openai interface is still accessed by default. Model name is empty. You need to select it manually every time:

[imoisture]

同样的问题：

[Issues-translate-bot]

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

---

Same question:

[imoisture]

@chunxingque 升级了一下docker镜像，好像又可以了~~~~~

环境变量配置如下

docker logs 查看到的运行日志如下

不过有个小问题，每次换浏览器访问，需要先**手动选择**一下 glm-4-plus (尽管启动容器的时候设置了只有这一个模型），才能正常输出。

[Issues-translate-bot]

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

---

@chunxingque upgraded the docker image and it seems to be working again~~~~~

The environment variables are configured as follows

docker logs The running logs viewed are as follows

However, there is a small problem. Every time you change the browser to access, you need to manually select glm-4-plus (even though only this model is set when starting the container) before it can output normally.