# Speech Recognition
URL: /docs/native/expo/speech-recognition

Setting up expo-speech-recognition for useSpeechRecognition and the DocyrusAgent microphone button.

## Overview

[`useSpeechRecognition`](/docs/native/hooks/use-speech-recognition) and the DocyrusAgent mic button (`DocyrusAgentChatInputMicButton` / `DocyrusAgentChatInputMic`) use `expo-speech-recognition`. It is an **optional peer**: without it (and always in **Expo Go**, which cannot load its native code) the hook reports `isSupported: false` and the mic button is hidden, so the rest of the chat keeps working.

## Installation

```bash
npx expo install expo-speech-recognition
```

**Dependencies:**
- [expo-speech-recognition](https://www.npmjs.com/package/expo-speech-recognition)

### Expo Configuration

Add the config plugin, then create a development build (`npx expo run:ios` / `eas build --profile development`):

```ts
{
  plugins: [
    [
      'expo-speech-recognition',
      {
        microphonePermission: 'Allow $(PRODUCT_NAME) to use the microphone for voice prompts.',
        speechRecognitionPermission: 'Allow $(PRODUCT_NAME) to transcribe your voice prompts.'
      }
    ]
  ]
}
```

The plugin writes `NSMicrophoneUsageDescription` / `NSSpeechRecognitionUsageDescription` on iOS and `RECORD_AUDIO` on Android. Permissions are requested on the first `start()`.

## Behaviour

| Situation | `isSupported` | Mic button |
|-----------|---------------|------------|
| Dev / production build with the module | `true` | Visible |
| Expo Go | `false` | Hidden |
| Module not installed | `false` | Hidden |
| Device reports recognition unavailable | `false` | Hidden |
| Permission denied | `true` (`error = 'not-allowed'`) | Visible, recording doesn't start |