Setup Google Generative AI provider and learn how to use its models like Gemini in the starter kit.
/ai/docs/google
Google AI
The Google Generative AI provider integrates Google's state-of-the-art models, including the versatile Gemini family, into your applications through the AI SDK.
Setup
### Generate API Key
Visit the [Google AI Studio](https://aistudio.google.com/app/apikey) to create your API key. For enterprise applications using Google Cloud, you can alternatively configure authentication via Application Default Credentials or service accounts.
### Add API Key to Environment
Add your API key to your project's `.env` file (e.g., in `apps/web`):
```bash title=".env"
GOOGLE_GENERATIVE_AI_API_KEY=your-api-key
```
If using Google Cloud credentials instead, ensure they're properly configured in your environment.
### Configure Provider (Optional)
The starter kit automatically uses the `GOOGLE_GENERATIVE_AI_API_KEY` environment variable. For advanced configurations (such as proxies, custom API versions, or specific headers), you can create a tailored provider instance using `createGoogleGenerativeAI`. See the [AI SDK Google documentation](https://sdk.vercel.ai/providers/ai-sdk-providers/google-generative-ai#provider-instance) for comprehensive details.
Features
Leverage Google's advanced Gemini models for chat, text generation,
reasoning, and complex instruction following.
Utilize text embedding models to convert text into numerical representations
for tasks like semantic search, clustering, and RAG.
Analyze and understand various file types (including images and PDFs)
alongside text prompts, enabling rich multimodal applications with
comprehensive content understanding.
Empower models to interact seamlessly with external tools and APIs, allowing
them to perform real-world actions and retrieve up-to-date information for
more capable applications.
Configure safety thresholds to control model responses regarding harmful
content categories. Access safety ratings in the response metadata.
Cache content to optimize context reuse and potentially reduce latency and
costs for repeated queries with similar context.
(With compatible models) Ground responses in real-time search results,
dramatically enhancing factual accuracy and providing up-to-date information
on current topics.
Use Cases
Create sophisticated conversational agents powered by Gemini models that can
engage in natural dialogue and handle complex, multi-step tasks. Experience
this in our [Chat Demo](/ai/docs/chat).
Generate diverse text formats, from creative writing and marketing copy to
code explanations and summaries.
Build applications that seamlessly analyze and understand images, documents,
and other file types alongside text, creating richer, more contextual user
experiences.
Implement powerful search capabilities or sophisticated Retrieval-Augmented
Generation systems using Google's high-performance embedding models for more
accurate information retrieval.
Streamline operations by connecting language models to external tools and
APIs through function calling, automating complex business processes and
repetitive tasks with minimal human intervention.
