Session isolation and codex model configuration.

[MaxMiksa]

A quick question regarding session isolation and model configuration.

My Scenario:

• I am currently running Project A using the medium model.
• I want to start a new Project B and switch it to use the xhigh model.

The Observation:

• I noticed that when I switch the model in the settings, the global configuration file on my C: drive updates immediately to reflect the new model parameter.

My Question:

• If I switch to xhigh for Project B, will my currently running Project A automatically reload the config and switch to xhigh as well?
• Or, does the running session load the config into memory only at startup (locking the medium model for the duration of the session), making it safe to run two different models in parallel?

I want to make sure I don't accidentally disrupt my running task by changing the global settings.

Thanks in advance!

[etraut-openai]

I presume you're using the CLI, as opposed to the IDE Extension.

The value stored in your config file informs which model is used when you create the next session. After the session is created, subsequent changes to the config file don't affect it.

[MaxMiksa]

Thank you for your response!
One quick follow-up: Will this logic apply to the vscode codex extension?

[etraut-openai]

The VS Code extension is a little different in that it doesn't write back changes to the config file when you change the model in its model picker. Like the CLI, when you start a new session in the IDE extension, the model is chosen based on your current model selection, and it won't change if you choose a different model in a different session (or if you edit your config file).

[MaxMiksa]

Thank you so much, truly helpful!