[Issue] Conversation fails after export while works well in dev.

[linkedlist771]

Summary

A short description of the issue.

Impact

• Blocks shipping a release
• Blocks development work
• Minor inconvenience

Steps To Reproduce

1. Install the release dmg file from release page.
2. Configure the API key and baseurl.
3. Start any converstaion

Expected Behavior

Tell us what you thought would happen.
We would expect it gives a response.

Actual Behavior

Describe what actually happened instead.
Here is a 401 authen error.

Screenshots Or Video

If available, attach screenshots, screen recordings, or GIFs that illustrate the issue.

Environment

• OS and version: Darwin
• Codexia version: 0.3.0
• Codex CLI version: 0.31
• Bun version: None(using the release version)
• Node.js version (if relevant):None(using the release version)
• Other tooling versions:None(using the release version)

Logs

Paste relevant terminal output, app logs, or link to a log snippet.

Additional Context

Anything else we should know?

[milisp]

If you use OpenAI, you can run codex at terminal then select ChatGPT to login, don't set api key or base_url for OpenAI

URL and api key is for other providers

[linkedlist771]

In the terminal, my codex keeps login, but in the codexia program, it still fails to stream, and gives the 401 error.:

I think this is caused by the envs passing, which seems not passed to the codex-cli in the codexia program.

[milisp]

Can you upgrade Codex CLI and try again, it's 0.40.0 now

Which model and provider you select?

[linkedlist771]

I have upgraded my codex-cli into 0.41.0, which turns out that in the terminal it works. If I just start the conversation using the GUI, its output stucked, but once a start a new conversation in terminal, the GUI's conversation magically works.

As the for the model, I use the gpt-5-codex, the provider I use a third party api which is openai compatiable.

[milisp]

It's magic.

I should create a custom provider.

[1Z-A11Y]

大佬你好，尝试example.py，下一步是什么？我用python运行不了

[milisp]

@1Z-A11Y 请问你用的什么版本？具体问题是什么？提供多点详情，如果你无法自己描述清晰，让 ChatGPT 帮你整理问题。实在不行请提供截图？什么系统？使用 OpenAI 还是别的提供商？如果是别的提供商就需要配置一下。

我几个小时前已经发布新版 0.7.1 ，可以定制其它提供商的 model 了，这个版本需要在配置文件 ~/.codex/config.toml 配置一下，具体配置参考 https://github.com/milisp/codexia/blob/main/docs/config.toml

然后在软件上配置一下 API key

What version are you using? What exactly is the issue? Provide more details. If you can't describe it clearly yourself, ask ChatGPT help you organize the problem. If that doesn't work, please provide screenshots? What operating system? Are you using OpenAI or another provider? If it's another provider, configuration is required.

I released version 0.7.1 a few hours ago, which now supports custom models from other vendors. This version requires configuration in the ~/.codex/config.toml file. For specific settings, please refer to: https://github.com/milisp/codexia/blob/main/docs/config.toml

Then configure the API key within the software.