RunAnywhere AI — iOS Example

A production-ready reference app for the RunAnywhere Swift SDK. LLM chat, speech, vision, voice agents, RAG, benchmarks, and model management—privacy-first and offline-capable on iPhone, iPad, and Mac.

Requirements

Item	Minimum
Xcode	26+ with Swift 6.2 and iOS 17.5+ simulator runtimes
Command Line Tools	Selected in Xcode → Settings → Locations
CMake & Ninja	For root native XCFramework generation
Disk space	Several GB for XCFramework output and AI models
Device	Apple Silicon recommended (physical device for MLX and best LLM performance)

Setup

Important: This sample links the local Swift SDK through Package.swift. A clean clone must build iOS XCFrameworks before Xcode can link the native backends.

1. Clone and enter the example

bash

git clone https://github.com/RunanywhereAI/runanywhere-sdks.git
cd runanywhere-sdks/examples/ios/RunAnywhereAI

2. Build native XCFrameworks (repo root)

bash

cd ../../..
./sdk/runanywhere-swift/scripts/build-core-xcframework.sh
cd examples/ios/RunAnywhereAI

Expected artifacts under sdk/runanywhere-swift/Binaries/:

Artifact	Needed for
`RACommons.xcframework`	Core (required)
`RABackendLLAMACPP.xcframework`	LLM and VLM
`RABackendONNX.xcframework` · `RABackendSherpa.xcframework`	STT, TTS, VAD
`onnxruntime.xcframework` · `onnx.xcframework`	ONNX Runtime for the two above
`RABackendMLX.xcframework` · `RunAnywhereMLXRuntime.xcframework` · `RunAnywhereMLXMetal.xcframework`	Apple MLX (this app links `RunAnywhereMLX`)
`RABackendCoreML.xcframework`	Core ML image generation

All ten are produced by one run of the script. A short list is the usual cause of link errors on the MLX path, since this app's Package.swift depends on the RunAnywhereMLX product.

Re-run this step after any C++ change in runanywhere-commons.

3. Resolve packages and build

bash

RUNANYWHERE_USE_LOCAL_NATIVES=1 swift package resolve

xcodebuild \
  -project RunAnywhereAI.xcodeproj \
  -scheme RunAnywhereAI \
  -resolvePackageDependencies

xcodebuild \
  -project RunAnywhereAI.xcodeproj \
  -scheme RunAnywhereAI \
  -configuration Debug \
  -sdk iphonesimulator \
  -destination 'generic/platform=iOS Simulator' \
  build

4. Run the app

Option A — Xcode: Open RunAnywhereAI.xcodeproj, select a simulator or device, press Run (⌘R).

Option B — Script:

bash

./scripts/build_and_run_ios_sample.sh simulator "iPhone 16 Pro" --build-sdk
# Physical device:
./scripts/build_and_run_ios_sample.sh device
# macOS:
./scripts/build_and_run_ios_sample.sh mac

Option C — Verify gate:

bash

./scripts/verify.sh

After modifying the SDK

Change	Action
Swift SDK source	Xcode picks up changes on rebuild
C++ / commons	Re-run `build-core-xcframework.sh`
Stale package errors	File → Packages → Reset Package Caches, then resolve again

Features

Feature	Description
AI Chat	Streaming LLM with thinking mode, tool calling, and LoRA adapters
Speech-to-Text	Batch and live transcription (Sherpa-ONNX / Whisper)
Text-to-Speech	Neural Piper voices
Voice Assistant	Full STT → LLM → TTS pipeline with particle UI
Vision (VLM)	Camera and photo-library image understanding
RAG	PDF/document ingestion and on-device Q&A
Benchmarks	Deterministic LLM, STT, TTS, and VLM performance tests
Voice Keyboard	iOS keyboard extension with dictation flow
Model Management	Download, load, storage, and deletion
Cross-Platform	Universal iOS, iPadOS, and macOS app

MLX-backed models run on physical iOS devices and native macOS; the arm64 simulator build validates package and startup paths but does not execute MLX inference.

Project structure

RunAnywhereAI/
├── RunAnywhereAI/
│   ├── App/                    # Entry point, SDK init, tab shell
│   ├── Features/               # Chat, Voice, Vision, RAG, Benchmarks, …
│   ├── Core/                   # Design system, services, models
│   └── Helpers/                # Markdown rendering, adaptive layout
├── RunAnywhereKeyboard/        # Keyboard extension target
├── RunAnywhereActivityExtension/  # Live Activity widget
├── Package.swift               # Local Swift SDK dependency
├── scripts/
│   ├── build_and_run_ios_sample.sh
│   ├── verify.sh
│   └── smoke.sh
└── README.md

Architecture: MVVM with Swift Observation (@Observable view models), a single RunAnywhere.* SDK entry point, and centralized design tokens (AppColors, AppTypography, brand orange #FF6900).

Troubleshooting

Symptom	Fix
Missing XCFramework errors	Run `./sdk/runanywhere-swift/scripts/build-core-xcframework.sh` from repo root
Package resolution failures	Reset package caches in Xcode; rerun `swift package resolve`
Sandbox / derived-data issues	Clean build folder (⇧⌘K); delete DerivedData if needed
MLX models unavailable	Use a physical device; MLX returns unavailable on simulator
Verify script fails	Run `./scripts/verify.sh` and follow its output for the exact gate

Quick static check without full compile:

bash

./scripts/smoke.sh

Filter runtime logs in Console.app: subsystem:com.runanywhere.RunAnywhereAI.

Resource	Link
Swift SDK	sdk/runanywhere-swift/README.md
Android example	examples/android/RunAnywhereAI
React Native example	examples/react-native/RunAnywhereAI
Flutter example	examples/flutter/RunAnywhereAI
App Store	RunAnywhere on the App Store
Discord	discord.gg/N359FBbDVd
Issues	GitHub Issues
Email	[email protected]

License

This project is licensed under the RunAnywhere License (Apache 2.0 based, with additional commercial-use terms). See LICENSE for details.

RunAnywhere AI — iOS Example

RunAnywhere AI — iOS Example

Requirements

Setup

1. Clone and enter the example

2. Build native XCFrameworks (repo root)

3. Resolve packages and build

4. Run the app

After modifying the SDK

Features

Project structure

Troubleshooting

Related links

License