React Native SDK (Android)

Package: teleneura-react-native

Thin bridge from JavaScript to the native com.teleneura.sdk:shared artifact. Batch accumulation, device metadata, and HTTPS ingestion run entirely on the native side — TypeScript only passes primitives (initialize, reportInference, flush).

Prerequisites

Android-only today — iOS would require a separate native implementation.
Maven Local: publish the Teleneura shared module from packages/teleneura-sdk-KMP:
```
cd packages/teleneura-sdk-KMP && ./gradlew :shared:publishToMavenLocal
```
Repositories: the React Native app’s Android Gradle config must include mavenLocal() alongside google() and mavenCentral().
Kotlin: shared is published with Kotlin 2.3.x. Align your host app’s Kotlin Gradle plugin (e.g. kotlinVersion = "2.3.0"). Older Kotlin toolchains cannot read newer library metadata.

Installation

From your application repo:

npm install /absolute/or/relative/path/to/packages/teleneura-rn-bridge

peerDependencies: React ≥ 18, React Native ≥ 0.74.

Autolinking is configured via react-native.config.js (TeleneuraPackage). After install:

cd android && ./gradlew clean && cd ..
npx react-native run-android

Manual link (only if autolinking is disabled): add TeleneuraPackage() alongside your other packages in MainApplication / ReactNativeHost getPackages().

import com.teleneura.rn.bridge.TeleneuraPackage

// Inside getPackages():
packages.add(TeleneuraPackage())

Usage

import { Teleneura } from 'teleneura-react-native'

await Teleneura.init({
  apiKey: process.env.TELENEURA_API_KEY!,
  baseUrl: 'https://api.example.com',
  batchSize: 32,
  enableDebugLogs: __DEV__,
})

// Fire-and-forget path — lowest bridge overhead per inference
Teleneura.reportInference('model-name', latencyMs)

// e.g. AppState → background
await Teleneura.flush()

The native module registers as TeleneuraReactNativeModule (NativeModules.TeleneuraReactNativeModule).

API

Method	Description
`init(config)`	`apiKey`, `baseUrl`, optional `batchSize`, `enableDebugLogs`. Resolves when native init completes.
`reportInference(modelName, latencyMs)`	Synchronous JS call; batches on native threads.
`flush()`	Sends buffered batches immediately.

Errors surface as promise rejections from init / flush with codes such as E_CONFIG, E_INIT, E_FLUSH.

Troubleshooting

Symptom	Check
Native module not found	Rebuild Android after `npm install`; confirm autolinking or manual `TeleneuraPackage`.
Gradle cannot resolve `shared`	Run `publishToMavenLocal`; verify `mavenLocal()` in repositories.
Kotlin metadata / incompatible version	Upgrade host Kotlin to 2.3.x (or publish `shared` with a Kotlin version matching your app — not recommended).