I have a Gemini key that works with the `gemini-3.5-flash` model. I am able to use this in `curl` and it works. In LiteLLM wizard config, it fails: ``` Testing connection to Google Gemini...

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

✘ Google Gemini — invalid API key ❯ Re-enter key? (y/N): ```

I would like debug this, so I tried `litellm --detailed_debug --setup` and `LITELLM_LOG=DEBUG litellm --setup` but neither gave more debug output.

I was able to use this successfully using the python client ``` python3 - <<'PY' import os import litellm

litellm._turn_on_debug()

response = litellm.completion( model="gemini/gemini-3.5-flash", messages=[{"role": "user", "content": "Reply with OK"}], api_key=os.environ["GEMINI_API_KEY"], )

print(response) PY ```

Why does this not work in the wizard?