1.3.2

Onde Inference Kotlin SDK 1.3.2

On-device LLM inference for Android and JVM (macOS Apple Silicon).

Add to your project

// build.gradle.kts
dependencies {
    implementation("com.ondeinference:onde-inference:1.3.2")
}

What's included

Target	Backend	Default model
Android (arm64-v8a, armeabi-v7a, x86_64, x86)	CPU (Candle)	Qwen 2.5 1.5B (~941 MB)
JVM (macOS Apple Silicon)	Metal	Qwen 2.5 3B (~1.93 GB)

Usage

// Android
val onde = OndeInference(applicationContext)

// JVM (macOS)
val onde = OndeInference()

// Shared API
val elapsed = onde.loadDefaultModel(systemPrompt = "You are helpful.")
val result  = onde.chat("Hello!")
onde.stream("Tell me a story.").collect { chunk -> print(chunk.delta) }

See the README for full API docs.

What's Changed

• Add ChatEngine::complete for caller-owned conversations by @setoelkahfi in #20

Full Changelog: 1.3.1...1.3.2