[Bug] 环境变量设置默认模型，前端UI不生效

[hamishliu]

📦 部署环境

Docker

📦 部署模式

服务端模式(lobe-chat-database 镜像)

📌 软件版本

v1.53.12

💻 系统环境

macOS

🌐 浏览器

Chrome

🐛 问题描述

设置了 DEFAULT_AGENT_CONFIG: "model=deepseek-chat" ，打开网页前端依然显示gpt-4o-mini

📷 复现步骤

设置 DEFAULT_AGENT_CONFIG: "model=deepseek-chat" 部署启动即可，打开网页

🚦 期望结果

显示环境变量配置的默认值

📝 补充信息

No response

[lobehubbot]

👀 @hamishliu

Thank you for raising an issue. We will investigate into the matter and get back to you as soon as possible.
Please make sure you have given us as much context as possible.
非常感谢您提交 issue。我们会尽快调查此事，并尽快回复您。 请确保您已经提供了尽可能多的背景信息。

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

📦 Deployment environment

Docker

📦 Deployment mode

Server-side mode (lobe-chat-database mirror)

📌 Software version

v1.53.12

💻 System environment

macOS

🌐 Browser

Chrome

🐛 Question description

DEFAULT_AGENT_CONFIG: "model=deepseek-chat" is set, and the front end of the web page still displays gpt-4o-mini

📷 Reproduction steps

Set DEFAULT_AGENT_CONFIG: "model=deepseek-chat" Deployment starts and opens the web page

🚦 Expected results

Display the default values ​​for environment variable configuration

📝 Supplementary information

No response

[dosubot[bot]]

@hamishliu 你好! 我是Dosu，一个可以帮你解决bug、回答问题，并成为贡献者的机器人。我会在这里帮助你，等待人类维护者的到来。祝你有愉快的一天! 🤖

A similar issue was reported where the front-end UI of LobeChat displayed "gpt-4o-mini" instead of the default model set in the DEFAULT_AGENT_CONFIG environment variable. This issue was resolved by ensuring that the DEFAULT_AGENT_CONFIG includes the correct provider configuration, such as provider=ollama, in addition to the model configuration ^[1]. Additionally, there was a closed issue where a pull request fixed the problem of the inbox agent not saving the configuration correctly ^[2].

If you are experiencing this issue in version 1.53.12, consider checking your configuration to ensure it includes the necessary provider details. If the problem persists, a page refresh might be required to apply the configured default model due to caching mechanisms in the front-end ^[3].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[cataglyphis]

+1

[noinlijin]

好像重启后第一次都会这样，刷新下就好了

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

It seems like this will happen for the first time after restarting, just refresh it

[2247ds]

清一下缓存就好了

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Just clear the cache

[MarchGe]

好像重启后第一次都会这样，刷新下就好了

不会好，问题一直存在，把enabled_openai和enabled_ollama关闭，顶部模型列表就出不来了，但是可以在设置里选择一个模型，系统在设置里把模型提供商写死了，比如我设置千问的qwen-turbo-latest作为默认模型，在设置里会自动显示成“openai/qwen-turbo-latest”😂，到现在这个问题还没修复。

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

It seems like this will happen for the first time after restarting, just refresh it

It won't be good, the problem always exists. Turn off enabled_openai and enabled_ollama, and the top model list will not be available, but you can select a model in the settings. The system will write the model provider to death in the settings, for example, I set up Qianwen. As the default model, qwen-turbo-latest will automatically display as "openai/qwen-turbo-latest" in the settings. This problem has not been fixed yet.

[bbbugg]

好像重启后第一次都会这样，刷新下就好了

不会好，问题一直存在，把enabled_openai和enabled_ollama关闭，顶部模型列表就出不来了，但是可以在设置里选择一个模型，系统在设置里把模型提供商写死了，比如我设置千问的qwen-turbo-latest作为默认模型，在设置里会自动显示成“openai/qwen-turbo-latest”😂，到现在这个问题还没修复。

是的，他只会选择openai的供应商，然后调用环境里设置的默认模型。注册一个新账号登录，可以直接看到这个bug。
但是，系统助手里的默认模型的环境变量设置是没有问题的，这个对话默认模型配置改成系统助手一样的用法就好了。

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

It seems like this will happen for the first time after restarting, just refresh it

It won't be good, the problem always exists. Turn off enabled_openai and enabled_ollama, and the top model list will not be available, but you can select a model in the settings. The system will write the model provider to death in the settings, for example, I set up thousands of questions As the default model, the qwen-turbo-latest will automatically display as "openai/qwen-turbo-latest" in the settings. This problem has not been fixed yet.

Yes, he will only select the vendor of openai and then call the default model set in the environment. Register a new account to log in and you can directly see this bug.
However, there is no problem with setting the environment variables of the default model in the system assistant. Just change the default model configuration to the same usage as the system assistant.

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

It seems like this will happen for the first time after restarting, just refresh it

It won't be good, the problem always exists. Turn off enabled_openai and enabled_ollama, and the top model list will not be available, but you can select a model in the settings. The system will write the model provider to death in the settings, for example, I set up thousands of questions As the default model, the qwen-turbo-latest will automatically display as "openai/qwen-turbo-latest" in the settings. This problem has not been fixed yet.

I understand, you have to write this
DEFAULT_AGENT_CONFIG=provider=goo gle/model=gemini-2.0-flash-exp

[bbbugg]

好像重启后第一次都会这样，刷新下就好了

不会好，问题一直存在，把enabled_openai和enabled_ollama关闭，顶部模型列表就出不来了，但是可以在设置里选择一个模型，系统在设置里把模型提供商写死了，比如我设置千问的qwen-turbo-latest作为默认模型，在设置里会自动显示成“openai/qwen-turbo-latest”😂，到现在这个问题还没修复。

我知道了，你要这样设置DEFAULT_AGENT_CONFIG=provider=google;model=gemini-2.0-flash-exp

[lobehubbot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

It seems like this will happen for the first time after restarting, just refresh it

It won't be good, the problem always exists. Turn off enabled_openai and enabled_ollama, and the top model list will not be available, but you can select a model in the settings. The system will write the model provider to death in the settings, for example, I set up thousands of questions As the default model, the qwen-turbo-latest will automatically display as "openai/qwen-turbo-latest" in the settings. This problem has not been fixed yet.

I understand, you need to set DEFAULT_AGENT_CONFIG=provider=google;model=gemini-2.0-flash-exp