Codex not see settings file

Resolved 💬 2 comments Opened Apr 29, 2025 by gitmeatarru Closed Aug 7, 2025

What version of Codex is running?

0.1.2504251709

Which model were you using?

qwen3
(LM Studio)

What platform is your computer?

windows 11 x64

What steps can reproduce the bug?

as I can see, you have finally added the ability to add other models, but this option does not seem to work yet.

as you have written in the documentation, you need to specify the following in the settings file::
provider_API_KEY
provider_BASE_URL

well, let's check it out, here's my file:

QWEN_API_KEY=""
QWEN_BASE_URL="http://192.168.0.12:1234"
model: qwen3-30b-a3b # Default model
fullAutoErrorMode: ask-user # or ignore-and-continue
notify: true # Enable desktop notifications for responses
safeCommands:
  - npm test

as you can see, it does not even see the specified model, although before that, there was gpt-4.1-nano and it saw everything perfectly!

The second attempt to force it to use the provider
--provider qwen
you can see in the screenshot below:
https://prnt.sc/YIafV2a3FbkE

For some reason, he decided that the model was o4 mini and that the provider was openai!!!
Why did you write me a ban by country, if I use a local model! lol
Or is it so fashionable now? That even a local network is important in which country?! It's funny!

What is the expected behavior?

execution of instructions

What do you see instead?

╭───────────────────────────────────
│ localhost session: 9f57e10eb6864b8d95e26d24d0181b5d
│ ↳ workdir: F:\DEV\PHP
│ ↳ model: o4-mini
│ ↳ provider: qwen
│ ↳ approval: suggest
╰───────────────────────────────────
user
make hellow worl app in html

system
⚠️ OpenAI rejected the request. Error details: Status: 403, Code: unsupported_country_region_territory, Type:
request_forbidden, Message: 403 Country, region, or territory not supported. Please verify your settings and try again.

Additional information

_No response_

View original on GitHub ↗

2 Comments

thisago · 1 year ago

Hello, is this your YAML config file?

QWEN_API_KEY=""
QWEN_BASE_URL="http://192.168.0.12:1234"
model: qwen3-30b-a3b # Default model
fullAutoErrorMode: ask-user # or ignore-and-continue
notify: true # Enable desktop notifications for responses
safeCommands:
  - npm test

If it is, you're adding Bash variables into YAML.

Move out the envs QWEN_API_KEY and QWEN_BASE_URL to your .bashrc (or any auto load file of your shell) with export keyword or just run in your terminal (in Windows I think it should be in Git bash or WSL):

export QWEN_API_KEY=""
export QWEN_BASE_URL=""

And it should work.

As it's not a good practice to save envs into config files, you keep them in your environment variables.

To make sure if it's available after exporting, you should be able to access the variables by echoing in terminal:

echo $QWEN_BASE_URL

codex-maintainers · 1 year ago

Thanks for the feedback! Please try again on the latest version of Codex CLI. If the issue persists, we'd love to hear about it.