No, you do not need to connect to Vertex AI to use Google models in the LLM class, but you must use the correct provider name. The provider should be "google-genai" (not "google") for Gemini models. Using "google" as the provider will result in an "unrecognized provider" error, as only specific provider names are supported by the LLM abstraction in Phoenix.
To use Gemini models, set the provider to "google-genai" and ensure you have the required dependencies (e.g., google-genai> =1.0.0). Example:
```
from phoenix.evals.llm import LLM
llm = LLM(provider="google-genai", model="gemini-2.5-flash")
```
See <https://github.com/Arize-ai/phoenix/blob/main/docs/evaluation/how-to-evals/configuring-the-llm.md|Configuring the LLM> and <https://arize.com/docs/phoenix/integrations/llm-providers/google-gen-ai/google-gen-ai-evals|Google Gen AI Evals> for details.
Would you like more detail on authentication or usage?
Sources:
- <https://arize.com/docs/phoenix/integrations/llm-providers/google-gen-ai/google-gen-ai-evals|Google Gen AI Evals>
- <https://github.com/Arize-ai/phoenix/blob/main/docs/evaluation/how-to-evals/configuring-the-llm.md|Configuring the LLM>
