Skip to main content
Read through this guide to set up Google Gemini & get familiar with how it is used across the different apps.

​
Set up Google Gemini

First, go to Google AI Studio and sign in with your Google account. Next, navigate to the API keys section and create a new API key. Make sure to save this somewhere safe and do not share it with anyone. Once you have your API key, paste it in your .env file: 
GOOGLE_GENERATIVE_AI_API_KEY=your_google_api_key

​
Available Models

Google Gemini models are defined in the unified model registry at lib/ai/models.ts.
ModelIDFeatures
Gemini 3 Progemini-3-pro-previewVision, Search Grounding
Gemini 3 Pro Imagegemini-3-pro-image-previewVision, Search Grounding, Image Generation
Gemini 3.1 Flash Litegemini-3.1-flash-lite-previewVision, Search Grounding
Gemini 2.5 Flashgemini-2.5-flashVision, Search Grounding, Thinking/Reasoning
Gemini models support search grounding, which enables native web search capabilities. This allows the model to access up-to-date information directly from the web when answering questions.
Gemini 2.5 Flash supports thinking capability, allowing the model to show its chain of thought before providing a final answer.

​
Apps Using Google Gemini

Google Gemini is integrated through the Vercel AI SDK 6.0. Provider routing is handled by lib/ai/ai-utils.ts using customModel(), getProviderFromModelId(), and getModelInstance().

Chat

Multi-provider chat app — Gemini is available as an LLM provider with search grounding support

Image Studio

Generate images using the Gemini 3 Pro Image model

Structured Output

Generate structured JSON output using Gemini models

​
Structure

The codebase uses Vercel AI SDK 6.0 with a unified model registry. Models are registered in lib/ai/models.ts and routing is managed through lib/ai/ai-utils.ts. The typical flow for an AI request:
  1. The user selects a model from the unified model registry
  2. The request is routed through getModelInstance() in lib/ai/ai-utils.ts
  3. The provider is determined via getProviderFromModelId()
  4. The model is instantiated with customModel()
  5. The response is streamed back to the user
  6. Results are stored in Supabase
More information on structure of the codebase can be found here:

Structure

Understand the project structure of the codebase