For Developers

On-Device AI SDK

Build private, offline-capable AI directly into your apps. Our SDK gives you the power of large language models that run entirely on-device — no cloud, no latency, no data leaks.

Get SDK Access Read Docs

Quick Start

Ship Private AI in Minutes

Initialize the SDK, load a model, and run inference in just a few lines of code. It really is that simple.

app.js

// Import the PocketAI SDK
import { PocketAI } from '@pocketai/sdk';

// Initialize with your configuration
const ai = new PocketAI({
  model:   'pocket-3b-instruct',
  device:  'auto',        // CPU, GPU, or NPU
  threads: 4,
  privacy: 'strict'       // Zero telemetry
});

// Run on-device inference
const response = await ai.generate({
  prompt: 'Summarize this medical record',
  context: patientData,
  maxTokens: 512,
  temperature: 0.3
});

console.log(response.text);
// => "Patient presents with..."
// All processing happened on-device

SDK Features

Everything You Need to Build

A complete toolkit for integrating on-device AI into iOS, Android, and cross-platform applications.

iOS & Android SDK

Native SDKs for both platforms with Swift and Kotlin APIs. Optimized for Apple Neural Engine, Qualcomm AI Engine, and MediaTek APU. Framework-agnostic with React Native and Flutter bindings available.

Custom Model Fine-tuning

Fine-tune base models on your own data with our training pipeline. LoRA and QLoRA support for efficient adaptation. Export optimized GGUF or CoreML models ready for on-device deployment.

API & Documentation

Comprehensive API reference, integration guides, and sample projects. Interactive playground for testing models before deployment. Full TypeDoc, Jazzy, and Dokka-generated references.

Apps Studio

Visual drag-and-drop interface for prototyping AI workflows without writing code. Test prompts, chain models, and export production-ready configurations directly to your app project.

Integration

How to Integrate in 3 Steps

From zero to on-device AI inference in under five minutes. No server infrastructure required.

Install the SDK

Add PocketAI to your project with your preferred package manager. Available on npm, CocoaPods, Maven, and pub.dev.

npm install @pocketai/sdk

Initialize

Configure the SDK with your API key and model preferences. The SDK automatically detects hardware capabilities and optimizes accordingly.

const ai = new PocketAI(config)

Run Inference

Call generate, classify, embed, or any task method. All processing happens on-device with zero network requests. Ship it.

await ai.generate({ prompt })