[Feature Request] 在 Codex model catalog 中实现 per-model reasoning levels 功能

[Curious-r]

桌面版近两个月加入、改进了这个功能，非常实用，可以解决在 codex 中使用第三方供应商时 reasoning levels 名字不一致的问题，不知是否有望在本项目中引进

谢谢！

[SaladDay]

Per-model reasoning levels and defaults for the Codex model catalog will be available in the next release.

[Curious-r]

Per-model reasoning levels and defaults for the Codex model catalog will be available in the next release.

One more thing I noticed that might be relevant here:

#442 looks like it nicely solves the per-model reasoning levels part I originally brought up in this issue — thanks for that!

I was wondering about OpenCode Go, though. The desktop version later had #6123 to handle the opencode.ai/zen/go reasoning behavior, since it needs to use the Zen gateway's reasoning format rather than the usual model/vendor-specific logic, and then use the per-model reasoningLevels from the catalog for clamping.

Not sure if the CLI already covers this part. If not, it might be worth keeping OpenCode Go in mind as a follow-up.

[SaladDay]

This hasn't been done yet, hope you can raise a follow-up issue