Wrong provider calls, fallbacking to Gemni CLI

[mihazs]

Description

Hello guys,

I've been using kilocode with VSCode LM API and gpt-4.1.

For some reason i got some strange rate limit with GPT-4.1 which is unlimited with GH Copilot pro.

I've investigate, and i was getting this error, telling about gemini cli:

I searched through the code, and this message only exists for gemini and qwen3. When I removed the configuration for gemini CLI, the issue has gone. In someway, kilocode are calling the wrong provider (gemini cli) instead of calling the VSCode LM API. Even if the selected provider right (vscode lm api)

[kevinvandijk]

Hi @mihazs,

Did you have gemini CLI configured in a separate profile and did it fall back to that profile or did the provider in your current profile change?

[mihazs]

I had in a separated profile, only worked when I deleted

[breixopd]

I'm having a similar issue even when the selected provider works. It will constantly try fall back to the copilot one even though I don't have a profile set up for it. The only profile I have set up is for Qwen coder and it will randomly start doing that and I have to disable and enable the extension again.