What causes "Gemini API key is not configured" in invoke-ai/InvokeAI?

Set the external_gemini_api_key config value (or its corresponding environment variable) to a valid Google AI Studio API key and restart InvokeAI.

invoke-ai/InvokeAI · error · ExternalProviderRequestError

Gemini API key is not configured

Error message

Gemini API key is not configured

What it means

ExternalProviderRequestError raised by GeminiProvider.generate when the app config has no external_gemini_api_key set. The provider requires an API key to authenticate requests to the Gemini generateContent endpoint (sent as the ?key= query parameter), and refuses to make a doomed request when it is absent. This is a configuration error, not a network or API failure.

Source

Thrown at invokeai/app/services/external_generation/providers/gemini.py:32

)
from invokeai.app.services.external_generation.image_utils import decode_image_base64, encode_image_base64


class GeminiProvider(ExternalProvider):
    provider_id = "gemini"
    _SYSTEM_INSTRUCTION = (
        "You are an image generation model. Always respond with an image based on the user's prompt. "
        "Do not return text-only responses. If the user input is not an edit instruction, "
        "interpret it as a request to create a new image."
    )

    def is_configured(self) -> bool:
        return bool(self._app_config.external_gemini_api_key)

    def generate(self, request: ExternalGenerationRequest) -> ExternalGenerationResult:
        api_key = self._app_config.external_gemini_api_key
        if not api_key:
            raise ExternalProviderRequestError("Gemini API key is not configured")

        base_url = (self._app_config.external_gemini_base_url or "https://generativelanguage.googleapis.com").rstrip(
            "/"
        )
        if not base_url.endswith("/v1") and not base_url.endswith("/v1beta"):
            base_url = f"{base_url}/v1beta"
        model_id = request.model.provider_model_id.removeprefix("models/")
        endpoint = f"{base_url}/models/{model_id}:generateContent"

        request_parts: list[dict[str, object]] = []

        if request.init_image is not None:
            request_parts.append(
                {
                    "inlineData": {
                        "mimeType": "image/png",
                        "data": encode_image_base64(request.init_image),
                    }

View on GitHub (pinned to 0b6a024f2f)

Solutions

Set the external_gemini_api_key config value (or its corresponding environment variable) to a valid Google AI Studio API key and restart InvokeAI.
Before selecting a Gemini model, call provider.is_configured() and surface a setup prompt to the user if it returns False.
Double-check the key is placed in the active config file/profile actually loaded by the app, not a stale one.
Verify the key works with a direct curl to https://generativelanguage.googleapis.com/v1beta/models?key=... .

Example fix

// before: key missing, generate() raises at runtime
result = gemini_provider.generate(request)
// after: check configuration first and give an actionable message
if not gemini_provider.is_configured():
    raise RuntimeError(
        "Gemini is not configured: set external_gemini_api_key in InvokeAI config"
    )
result = gemini_provider.generate(request)

Defensive patterns

Strategy: validation

Validate before calling

if not app_config.external_gemini_api_key:
    raise RuntimeError("Gemini is not configured: set external_gemini_api_key before using Gemini models")
# or, non-raising:
if not gemini_provider.is_configured():
    disable_gemini_models_in_ui()

Try / catch

try:
    result = provider.generate(request)
except ExternalProviderRequestError as e:
    if "not configured" in str(e):
        show_setup_instructions("gemini")
    else:
        raise

Prevention

Check provider.is_configured() at startup and expose the result in the UI/model list.
Keep API keys in environment variables loaded by the actual service process.
Add a config smoke test that asserts required keys exist when a provider is enabled.
Document which env vars each provider needs.

When it happens

Trigger: A generation request is routed to provider_id 'gemini' while self._app_config.external_gemini_api_key is empty/None — i.e., generate() is called without the key ever having been configured (is_configured() would also return False).

Common situations: Missing environment variable / config entry for the Gemini API key; key set under the wrong config name (e.g. only OPENAI key configured); key removed after a config migration; user selects a Gemini model without ever enrolling an API key.

Understand the failure class

Background: "environment variable is not set" and "Missing keys in environment" errors: what missing required env var messages mean and how to fix them — this error's family across 28 libraries.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/bdbe27ffebe1a7d1. Report an issue: GitHub.