ContextQMD
Back to Eliza

Google Gemini Plugin

packages/docs/plugin-registry/llm/google-genai.md

2.0.13.4 KB
Original Source

The Google Gemini plugin connects Eliza agents to Google's Gemini API, providing access to the Gemini 2.5 and 2.0 model families with multimodal input support.

Package: @elizaos/plugin-google-genai

Installation

bash
eliza plugins install @elizaos/plugin-google-genai

Auto-Enable

The plugin auto-enables when GOOGLE_GENERATIVE_AI_API_KEY or GOOGLE_API_KEY is present:

bash
export GOOGLE_GENERATIVE_AI_API_KEY=AIza...
# or
export GOOGLE_API_KEY=AIza...

Configuration

Environment VariableRequiredDescription
GOOGLE_GENERATIVE_AI_API_KEYYes*Google AI Studio API key
GOOGLE_API_KEYYes*Alias (also triggers auto-enable)
GOOGLE_SMALL_MODELNoOverride the small model identifier
GOOGLE_LARGE_MODELNoOverride the large model identifier
SMALL_MODELNoGlobal alias to override the small model
LARGE_MODELNoGlobal alias to override the large model
GOOGLE_EMBEDDING_MODELNoOverride the embedding model identifier
GOOGLE_IMAGE_MODELNoOverride the image generation model
IMAGE_MODELNoGlobal alias for the image model

*Either GOOGLE_GENERATIVE_AI_API_KEY or GOOGLE_API_KEY is accepted.

Get your API key from aistudio.google.com.

eliza.json Example

json
{
  "auth": {
    "profiles": {
      "default": {
        "provider": "google-genai",
        "model": "gemini-2.5-pro"
      }
    }
  }
}

Supported Models

Gemini 2.5 Family

ModelContextBest For
gemini-2.5-pro1M tokensComplex reasoning, long context
gemini-2.5-flash1M tokensFast, cost-efficient tasks
gemini-2.5-flash-lite1M tokensHighest throughput

Gemini 2.0 Family

ModelContextBest For
gemini-2.0-flash1M tokensBalanced performance
gemini-2.0-flash-lite1M tokensEfficient tasks

Gemini 1.5 Family (Legacy)

ModelContextBest For
gemini-1.5-pro2M tokensMaximum context window
gemini-1.5-flash1M tokensCost-efficient

Model Type Mapping

elizaOS Model TypeGemini Model
TEXT_SMALLgemini-2.0-flash-001
TEXT_LARGEgemini-2.0-flash-001
TEXT_EMBEDDINGtext-embedding-004
IMAGE_DESCRIPTIONgemini-2.0-flash-001 (vision)

Features

  • Streaming responses
  • Function calling / tool use
  • Vision (images, video, audio input)
  • 1M+ token context windows (Gemini 2.5)
  • Code execution capability
  • Structured JSON output
  • Grounding with Google Search (Gemini 2.0+)

Multimodal Input

Gemini models natively accept images, audio, video, and documents as input:

typescript
const response = await runtime.useModel("IMAGE_DESCRIPTION", {
  imageUrl: "https://example.com/chart.png",
  prompt: "Describe what this chart shows.",
});

Rate Limits and Pricing

Free tier: Available via Google AI Studio (rate-limited). Paid tier: Via Google Cloud Vertex AI or Google AI Studio billing.

See ai.google.dev/pricing for current rates.