# @standardagents/google

Google Gemini and Imagen provider package for Standard Agents using the official @google/genai SDK

> **Info:**
>
> For direct access, set `GOOGLE_API_KEY` in your environment. Standard Agents Builder can use hosted routing through `STANDARD_AGENTS_API_KEY` when the provider key is absent. Hosted access depends on platform configuration.

## Overview

The `@standardagents/google` package provides a Standard Agents provider for Google's Gemini and Imagen APIs. It uses the official `@google/genai` SDK, supports Gemini text/tool/image requests, and routes Imagen generation and editing through Google's dedicated image endpoints.

- Key Features

  - Official `@google/genai` SDK integration
  - Gemini text, multimodal, JSON, streaming, and tool calling support
  - Gemini thought-signature preservation for multi-step tool calling
  - Imagen image generation and edit support
  - Built-in pricing for documented Gemini models used in request logs
  - Public model discovery through `models.list()`

## Installation

```bash pnpm
pnpm add @standardagents/google
```

## Quick Start

```typescript
import { defineModel } from '@standardagents/spec';
import { google } from '@standardagents/google';

export default defineModel({
  name: 'gemini-fast',
  provider: google,
  model: 'gemini-3.8-flash',
  providerOptions: {
    responseModalities: ['TEXT'],
  },
});
```

## Common Models

- `gemini-3.1-pro-preview`
- `gemini-3.8-flash`
- `gemini-2.5-flash-lite`
- `gemini-2.5-flash-image`
- `gemini-3-pro-image-preview`
- `imagen-4.0-generate-001`
- `imagen-3.0-capability-001`

## Provider Options

Common Gemini options include `responseModalities`, `candidateCount`, `cachedContent`, `thinkingConfig`, `imageConfig`, and `safetySettings`.

Common Imagen options include `numberOfImages`, `negativePrompt`, `aspectRatio`, `outputMimeType`, `imageSize`, and `editMode`.

## Tool Result Images

The Google provider preserves Standard Agents' internal tool-result message structure and only rewrites the outbound Google request:

- Gemini 3 series models keep image tool results nested in `functionResponse.parts`
- Older Google chat models get a text-only `functionResponse`, with the image emitted as a sibling inline image part instead

This keeps provider fallbacks and log storage provider-agnostic while still working around older Google models that reject multimodal `functionResponse` payloads.

## Pricing

The provider includes built-in pricing for documented Gemini models such as `gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.5-flash-lite`, `gemini-2.5-flash-image`, and `gemini-3-pro-image-preview`.

The provider also includes per-image pricing for Imagen 4 generation variants and `imagen-3.0-capability-001`. For models without bundled pricing, set prices on your model definition.

## Environment Setup

```bash .dev.vars
GOOGLE_API_KEY=...
```