Fallback to gemini-2.5-flash not respected. Requests still sent to gemini-2.5-pro

[blueraft]

TL;DR

Thanks for this, awesome tool!

When gemini-2.5-pro is temporarily unavailable, the system logs indicate that it switches to gemini-2.5-flash. However, actual API requests continue to be made to gemini-2.5-pro, causing quota errors.

Expected behavior

No response

Observed behavior

No response

Action YAML

The default yaml created by gemini-cli.

Log output

Run google-github-actions/run-gemini-cli@v0
.....
[INFO] Your configured model (gemini-2.5-pro) was temporarily unavailable. Switched to gemini-2.5-flash for this session.

Attempt 1 failed with status 429. Retrying with backoff...
ApiError: {
  "error": {
    "message": "{
      \"error\": {
        \"code\": 429,
        \"message\": \"You exceeded your current quota, please check your plan and billing details...\",
        \"status\": \"RESOURCE_EXHAUSTED\",
        ...
        \"quotaDimensions\": {
          \"model\": \"gemini-2.5-pro\",
          ...
        }
      }
    }"
  }
}


https://github.com/FAIRmat-NFDI/nomad-docs/actions/runs/16805741946/job/47597567647

Additional information

No response

[tatarysh]

I'm testing gemini-cli on github, after one task I exceeded the limit (640k tokens), I have no idea how it happened. of course, it did not switch to another model.

[jerop]

Routed upstream: google-gemini/gemini-cli#6742