BerriAI/litellm · error · ValueError

Google API key is required. Set GOOGLE_API_KEY or…

Error message

Google API key is required. Set GOOGLE_API_KEY or GEMINI_API_KEY environment variable.

What it means

Raised while building the URL for the Gemini Interactions API (POST /v1beta/interactions, used by litellm.interactions.create). GeminiModelInfo.get_api_key resolves litellm_params['api_key'], then the GOOGLE_API_KEY env var, then GEMINI_API_KEY; if all are empty the request is aborted before it starts. Note this path (unlike the image handlers) does accept GOOGLE_API_KEY.

Solutions

export GOOGLE_API_KEY=... or GEMINI_API_KEY=... before starting the process
Or pass api_key via litellm_params: litellm.interactions.create(..., api_key=KEY)
Confirm with: python -c "from litellm.llms.gemini.common_utils import GeminiModelInfo; print(bool(GeminiModelInfo.get_api_key()))"
In containers, verify the env var exists inside the running process (printenv)

Example fix

# before
r = litellm.interactions.create(model='gemini-2.5-flash', input={'text': 'hi'})
# ValueError: Google API key is required. Set GOOGLE_API_KEY or GEMINI_API_KEY environment variable.

# after
r = litellm.interactions.create(
    model='gemini-2.5-flash',
    input={'text': 'hi'},
    api_key=os.environ['GOOGLE_API_KEY'],
)

Defensive patterns

Strategy: validation

Validate before calling

from litellm.llms.gemini.common_utils import GeminiModelInfo

assert GeminiModelInfo.get_api_key(), \
    'interactions.create needs api_key or GOOGLE_API_KEY/GEMINI_API_KEY in env'

Prevention

Export GOOGLE_API_KEY/GEMINI_API_KEY in every service that creates interactions
Pass api_key explicitly from your secret store rather than trusting ambient env
Fail fast at startup with GeminiModelInfo.get_api_key() checks

When it happens

Trigger: litellm.interactions.create(model='gemini-2.5-flash', input=...) with no api_key in litellm_params and neither GOOGLE_API_KEY nor GEMINI_API_KEY in the environment; streaming creates hit the same check before the ?alt=sse URL is built.

Common situations: New Interactions-API integrations where env vars were wired for other providers only; serverless/cold-start processes missing env config; keys stored in a vault but not exported into the process.

Understand the failure class

Background: "API key is required" / "API key not found" / "No API key was set": the missing-api-key error family across 16 libraries — this error's family across 16 libraries.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/777981c528c26d78. Report an issue: GitHub.

Appendix: source

Thrown at litellm/llms/gemini/interactions/transformation.py:116

            headers["Api-Revision"] = "2026-05-20"

        return headers

    def get_complete_url(
        self,
        api_base: str | None,
        model: str | None,
        agent: str | None = None,
        litellm_params: dict | None = None,
        stream: bool | None = None,
    ) -> str:
        """POST /{api_version}/interactions"""
        litellm_params = litellm_params or {}
        api_base = GeminiModelInfo.get_api_base(api_base)
        api_key: Final = GeminiModelInfo.get_api_key(litellm_params.get("api_key"))

        if not api_key:
            raise ValueError("Google API key is required. Set GOOGLE_API_KEY or GEMINI_API_KEY environment variable.")

        if stream:
            return f"{api_base}/{self.api_version}/interactions?alt=sse"

        return f"{api_base}/{self.api_version}/interactions"

    def transform_request(
        self,
        model: str | None,
        agent: str | None,
        input: InteractionInput | None,
        optional_params: InteractionsAPIOptionalRequestParams,
        litellm_params: GenericLiteLLMParams,
        headers: dict,
    ) -> dict:
        """
        Build request body per OpenAPI spec.

View on GitHub (pinned to 77b7c6c40c)