nobodywho-ooo / nobodywho
A Rust-powered local LLM inference engine with unprecedented cross-platform bindings for game engines, mobile frameworks, and multiple languages
星标趋势
AI 分析
项目摘要
NobodyWho is a Rust-based inference engine that enables efficient local execution of LLMs on any device, with bindings for Python, Flutter, Godot, Kotlin, Swift, and React Native. It supports multimodal input, tool calling, text-to-speech, speech-to-text, and voice activity detection, all running offline without API keys.
为什么值得关注
NobodyWho stands out for its unique cross-platform approach, bringing local LLM inference to game engines (Godot), mobile frameworks (Flutter, React Native), and multiple programming languages through a single Rust core. Its integration of both text generation (via llama.cpp) and speech capabilities (via ONNX Runtime) in one unified engine is rare and makes it particularly valuable for on-device AI applications.
优势
- Exceptional cross-platform support with bindings for Python, Flutter, Godot, Kotlin, Swift, and React Native
- Comprehensive on-device AI capabilities including LLM inference, TTS, STT, and VAD in a single engine
- Fast, type-safe tool calling with automatic grammar generation from function signatures
局限性
- No test suite despite having CI, which raises concerns about code reliability
- EUPL-1.2 license may be less familiar and potentially restrictive for some commercial use cases compared to Apache or MIT
使用场景
- Running local LLMs in Godot games for NPCs with natural language interactions
- Building offline-capable mobile AI assistants with Flutter or React Native
- Creating privacy-preserving AI applications that process data entirely on-device