Models & Availability

adk() runs Gemini Nano on the device. Android provides it through AICore, so there’s nothing to download or bundle, but only some devices can run it. When you need a larger model, or support for every device, use cloud Gemini.

Model	`modelType`	`modelName`	Runs
Gemini Nano	`genai-nano`	`gemini-nano` (the default)	On device, offline, no API key
Cloud Gemini	`gemini`	Required, such as `gemini-2.5-flash`	Google AI API, with your API key

Gemini Nano

Availability Check

Check before you offer Gemini Nano in your app. There are two checks, both cheap and safe to run at startup and again when the app resumes:

Call	Answers
`adk.isNanoSupported()`	Can this device ever run Gemini Nano?
`adk.isAvailable('genai-nano')`	Can it run now, or is it still being downloaded?

import { adk } from '@react-native-ai/adk'

if (!(await adk.isNanoSupported())) {
  // Hide Gemini Nano: this device can't run it
  return
}

if (!(await adk.isAvailable('genai-nano'))) {
  // Show it as not ready yet, for example while Android downloads it
  return
}

await adk.prepareNano()
const model = adk()

Both are based on ML Kit’s model status. See the ML Kit GenAI documentation for the supported devices.

Preparing

adk.prepareNano() loads Gemini Nano ahead of time. If you skip it, the first generateText or streamText call prepares it. That call only checks isNanoSupported(), so check isAvailable('genai-nano') yourself before showing the model as ready, and handle errors in your chat flow.

Cloud Gemini

Use cloud Gemini on devices without Gemini Nano, on the emulator, or when you need a more capable model. Create a provider with modelType: 'gemini', a modelName and your API key:

import { createAdkProvider } from '@react-native-ai/adk'
import { generateText } from 'ai'

const adk = createAdkProvider({
  apiKey: process.env.GOOGLE_API_KEY,
  modelType: 'gemini',
  modelName: 'gemini-2.5-flash',
})

const { text } = await generateText({
  model: adk(),
  prompt: 'What time is it in New York?',
})

Cloud models need no download or prepare step.

Provider Options

createAdkProvider creates a provider with its own settings, for example a system instruction:

import { createAdkProvider } from '@react-native-ai/adk'

const adk = createAdkProvider({
  instruction: 'You are a concise, friendly assistant.',
})

const model = adk()

Option	Type	Description
`modelType`	`'genai-nano' \| 'gemini'`	On device or cloud (default: `genai-nano`)
`modelName`	`string`	Model name (default: `gemini-nano`; set it for cloud Gemini)
`apiKey`	`string`	Google AI API key, for cloud Gemini only
`instruction`	`string`	System instruction
`availableTools`	`Record<string, Tool>`	Tools ADK can call, see Tool Calling
`name`	`string`	ADK agent name (default: `react_native_adk_agent`)
`description`	`string`	ADK agent description