runanywhere-sdks

RunAnywhere is a set of SDKs for adding AI features to mobile apps that run entirely on-device, with no cloud, no network latency and no data leaving the device. It covers LLM chat (Llama, Mistral, Qwen, SmolLM and others), Whisper-based speech-to-text, neural text-to-speech and a full voice-assistant pipeline that chains speech recognition, an LLM and speech synthesis.

SDKs exist for Swift (iOS and macOS, stable, via Swift Package Manager), Kotlin (Android, stable, via Gradle), React Native (beta, npm) and Flutter (beta, pub.dev). The feature table lists text generation, streaming, speech-to-text, text-to-speech, the voice assistant pipeline and model download with progress on all four, with structured JSON output on iOS and Android and Apple Foundation Models on iOS only. Supported models include GGUF LLMs run through llama.cpp (from SmolLM2 360M at about 400 MB up to Mistral 7B Q4 at about 4 GB), Whisper Tiny and Base via ONNX, and Piper voices for US and British English. It is Apache 2.0 licensed.