Skip to main content

Google Provider

💡TL;DR

Google uses the native Gemini Generative Language API transport. The current preset defaults to gemini-2.0-flash-exp, https://generativelanguage.googleapis.com/v1, and temperature: 0.5. It is not an OpenAI-compatible profile.

This is part of the Obsidian AI Knowledge Management Guide.

Setup

Create a Google AI Studio or Google Cloud API key with Gemini access, add a Google provider profile, and keep the native Generative Language Base URL. Use the default Gemini Flash profile for fast translation and broad note workflows, then override per-task models when you need a larger or more capable Gemini model.

Avoid placing Gemini model IDs in an OpenAI-compatible profile unless you are explicitly using a gateway that exposes Gemini through an OpenAI-compatible API.

Endpoint And Authentication

FieldCurrent preset
Transportgoogle
API keyRequired
Base URLhttps://generativelanguage.googleapis.com/v1
Default modelgemini-2.0-flash-exp
Temperature0.5
Connection testChat-only

The runtime calls /models/{model}:generateContent?key={apiKey}. The key is passed as a query parameter because that is how the native Gemini endpoint is wired in the plugin.

Model Discovery

Google uses the google-models discovery mode. Discovery should return Gemini models available to the API key; if the selector is empty, test the key against the same Base URL and confirm the model family is enabled for that account.

Known output-token ceilings are model-aware in the provider registry. If a long translation or diagram explanation is truncated, select a Gemini model with a larger supported output cap or lower the requested output size.

Troubleshooting

  • 400: the model ID usually does not match the native Gemini route or the request exceeds the model's supported shape.
  • 401 or 403: verify key restrictions, project/API enablement, and whether the key can access Gemini.
  • 404: the model ID is not available at the configured API version.
  • Gateway examples fail under this profile: use the gateway preset instead of the native Google provider.

When To Use

Use Google for fast translation, large-volume note cleanup, and workflows where Gemini's cost/performance tradeoff is better than direct OpenAI or Anthropic. Prefer another provider for workflows that require OpenAI-compatible reasoning-model handling.


Next Steps

  • LLM Providers -- Full configuration, per-task models, and retry logic
  • Translation -- Gemini Flash is recommended for high-volume translation tasks