React Native · Nosmai Moderation Docs

Install

npm install @nosmai/moderation-react-native

The native SDK ships with the plugin:

iOS: run pod install. Requires iOS 14+.
Android: download nosmai-detection.aar from the releases page into android/app/libs/ and reference it, with minSdk 24, arm64-v8a, and Kotlin 2.2.0+.

Requirements: React Native with the New Architecture enabled. For the live camera add NSCameraUsageDescription (iOS) and the CAMERA permission (Android); on iOS also set ITSAppUsesNonExemptEncryption to NO in Info.plist.

Initialize

Call once at startup with your license key and the models you need. Native inference runs off the JS thread.

import { NosmaiModeration } from '@nosmai/moderation-react-native';

const init = await NosmaiModeration.initialize('NOSMAI-XXXX', [
  'objectDetection',
  'nsfw',
  'text',
]);
if (!init.success) console.warn('init failed:', init.error);

Moderate an image

Pass a file path (for example from react-native-image-picker; strip the file:// prefix).

const r = await NosmaiModeration.analyzeImage(filePath);   // NosmaiResult
if (r.isUnsafe) {
  // r.detections -> [{ category, confidence }], r.nsfw -> 'safe' | 'warn' | 'block'
}

Moderate a video

const v = await NosmaiModeration.analyzeVideo(filePath, 500);  // frameIntervalMs
if (v.isUnsafe) {
  // v.categories, v.framesAnalyzed, v.nsfw
}

Moderate text

Load the text model once (after initialize), then moderate messages.

await NosmaiModeration.initializeText();

const t = await NosmaiModeration.moderateText('some message');
if (t.blocked) {
  // t.layer ('blocklist' | 'classifier'), t.category, t.matchedWord
}

Live camera

Render <NosmaiCameraPreview />. The native view owns the camera: it starts on mount and stops on unmount. Request camera permission before mounting it.

import { NosmaiCameraPreview } from '@nosmai/moderation-react-native';

<NosmaiCameraPreview
  style={{ width: '100%', height: 320 }}
  facing="back"                       // or "front"
  onResult={(r) => {
    // r.isUnsafe, r.nsfw, r.detections — one result per analyzed frame
  }}
/>

On the live camera, a suggestive (warn) verdict that persists for several seconds is escalated to block, so sustained borderline content is treated as unsafe.

Thresholds

NosmaiModeration.setThreshold('weapon', 0.7);
NosmaiModeration.setNsfwThreshold('explicit', 0.45);  // block
NosmaiModeration.setNsfwThreshold('sexy', 0.55);      // warn

Cleanup

NosmaiModeration.shutdown();

[!NOTE] Enums cross the bridge as strings, so category, nsfw, layer and the rest are plain string values you can compare directly.