microsoft/semantic-kernel · error · ServiceInitializationError

The API key is required when use_vertexai is False.

Error message

The API key is required when use_vertexai is False.

What it means

Raised during GoogleAIChatCompletion.__init__ when use_vertexai is False (the default), no client was injected, and api_key is missing. In API-key mode the Gemini client authenticates with the key, so SK rejects initialization when it is absent.

Source

Thrown at python/semantic_kernel/connectors/ai/google/google_ai/services/google_ai_chat_completion.py:125

                cloud_project_id=project_id,
                cloud_region=region,
                use_vertexai=use_vertexai,
                env_file_path=env_file_path,
                env_file_encoding=env_file_encoding,
            )
        except ValidationError as e:
            raise ServiceInitializationError(f"Failed to validate Google AI settings: {e}") from e

        if not google_ai_settings.gemini_model_id:
            raise ServiceInitializationError("The Google AI Gemini model ID is required.")

        if not client:
            if google_ai_settings.use_vertexai and not google_ai_settings.cloud_project_id:
                raise ServiceInitializationError("Project ID must be provided when use_vertexai is True.")
            if google_ai_settings.use_vertexai and not google_ai_settings.cloud_region:
                raise ServiceInitializationError("Region must be provided when use_vertexai is True.")
            if not google_ai_settings.use_vertexai and not google_ai_settings.api_key:
                raise ServiceInitializationError("The API key is required when use_vertexai is False.")

        super().__init__(
            ai_model_id=google_ai_settings.gemini_model_id,
            service_id=service_id or google_ai_settings.gemini_model_id,
            service_settings=google_ai_settings,
            client=client,
        )

    # region Overriding base class methods

    # Override from AIServiceClientBase
    @override
    def get_prompt_execution_settings_class(self) -> type["PromptExecutionSettings"]:
        return GoogleAIChatPromptExecutionSettings

    @override
    @trace_chat_completion(GoogleAIBase.MODEL_PROVIDER_NAME)
    async def _inner_get_chat_message_contents(

View on GitHub (pinned to c028a0c7dc)

Solutions

1. Set GOOGLE_AI_API_KEY to your Gemini API key, or pass api_key='...' to the constructor.
2. If you meant to use Vertex AI, set use_vertexai=True and provide project_id and region instead.
3. Provide a prebuilt google.genai Client via client= to bypass the API-key requirement.

Example fix

# before
svc = GoogleAIChatCompletion(gemini_model_id="gemini-1.5-pro")  # no key
# after
svc = GoogleAIChatCompletion(gemini_model_id="gemini-1.5-pro", api_key="AIza...")

def api_key_present_when_not_vertex(use_vertexai: bool, api_key: str | None, client) -> bool:
    if not use_vertexai and client is None:
        return isinstance(api_key, str) and bool(api_key.strip())
    return True

def is_valid_api_key_config(use_vertexai: bool, api_key: object, has_client: bool) -> bool:
    return use_vertexai or has_client or (isinstance(api_key, str) and bool(api_key.strip()))

from semantic_kernel.exceptions.service_exceptions import ServiceInitializationError

try:
    svc = GoogleAIChatCompletion(gemini_model_id=mid, api_key=key)
except ServiceInitializationError as e:
    if "API key is required" in str(e):
        raise ValueError("Set GOOGLE_AI_API_KEY or pass api_key (or enable Vertex AI)") from e
    raise

When it happens

Trigger: Constructing with use_vertexai=False (default) without GOOGLE_AI_API_KEY or an api_key argument, and without providing a prebuilt client.

Common situations: First-time setup without an API key; env var typo GOOGLE_AI_API_KEY; .env not loaded; key present in a different env prefix; intending to use Vertex AI but use_vertexai still defaults to False.

Related errors

• The API key is required when use_vertexai is False.
• The API key is required when use_vertexai is False.
• Copilot Studio SERVICE authentication is not available yet.
• Please provide either an api_key, ad_token, ad_token_provide
• The OpenAI API key is required.

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/df0b58175a694353. Report an issue: GitHub.