Local, offline inference
Run models locally and offline without API keys or ongoing service fees, which keeps inference self-contained on the device.
NobodyWho
NobodyWho is an on-device inference engine for running LLMs locally without API keys or cloud calls. Supports multiple app frameworks and GGUF models from Hugging Face or any URL.
Overview
NobodyWho is an inference engine for running large language models locally and efficiently on supported devices. It is positioned as on-device AI for apps that need chat model inference without sending requests to an external API.
The README emphasizes offline operation, support for GGUF-format models, and compatibility with multiple client platforms including Kotlin, Swift, Python, Flutter, React Native, and Godot. It also documents paths for loading models from Hugging Face or a direct URL, along with support for GPU-accelerated inference through Vulkan or Metal.
Features
Broad model compatibility
Work with chat models including Gemma, Qwen, Mistral, and other GGUF-based LLMs, rather than being tied to a single model family.
Type-safe tool calling
Define function signatures and have the engine generate structured grammars for tool calls automatically, avoiding manual schema work.
Multimodal input support
Send image and audio information to the model, enabling multimodal prompts where supported by the host app.
Built-in voice workflows
Use local text-to-speech and speech-to-text components, including Kokoro, Pocket TTS, Supertonic, and Whisper.
Flexible model loading
Load models from Hugging Face or any URL, which makes it easier to point an app at a remote model file or download source.
Use Cases
Offline chat experiences
Build an app that answers user prompts on-device without sending data to a hosted LLM API, useful when you want offline operation or fewer external dependencies.
Tool-using app workflows
Add LLM-backed actions to an app by exposing functions from your code and letting NobodyWho generate the structured grammar for calls.
Cross-platform app development
Create mobile or desktop apps in supported stacks such as Swift, Flutter, React Native, Kotlin, or Python and connect them to the same local inference layer.
Voice-enabled assistants
Combine chat with local audio features, such as speech-to-text for transcription or text-to-speech for generated responses, in a single device-side workflow.
Custom model loading
Load a model from Hugging Face or a direct URL when you want to point the app at a specific model artifact instead of a hosted API endpoint.
Pros and Cons
Pros
- Runs locally and offline, which reduces dependency on cloud APIs.
- Supports several app development stacks, including Kotlin, Swift, Python, Flutter, React Native, and Godot.
- Works with models in GGUF format and can load from Hugging Face or any URL.
- Includes structured tool calling plus voice-related capabilities such as text-to-speech and speech-to-text.
- Repository licensing text says commercial and proprietary use is allowed free of charge, subject to the stated conditions on modifications.
Cons
- The source does not give detailed benchmarks, memory requirements, or supported device list, so performance expectations are still unclear.
- Some platform support details are summarized at a high level; feature parity and limits may differ by binding.
- Web export is described as more difficult to get right, with an issue referenced in the README.
FAQ
Does NobodyWho require an API key?
NobodyWho can run locally without API keys or hidden fees, so it is designed for offline use on supported devices and platforms.
Which platforms does it support?
The README lists Kotlin, Swift, Python, Flutter, React Native, and Godot, with documentation links for each.
What kinds of models can it use?
It supports chat LLMs such as Gemma, Qwen, Mistral, and other models in GGUF format, and it can load models from Hugging Face or a direct URL.
Can I use it on mobile?
The README says desktop is supported across all bindings, Android is supported on Kotlin, Godot, Flutter, and React Native, and iOS is supported on Swift, Flutter, and React Native.
Is NobodyWho free to use in commercial projects?
The repository license section says you may use NobodyWho in proprietary and commercial projects free of charge, but modified versions of the code in the repo must be open sourced.
Quick Facts
- Category
- Developer Tool / AI Inference
- Source domain
- github.com
- Primary use
- Local LLM inference on device
- Supported platforms
- Kotlin, Swift, Python, Flutter, React Native, Godot
- Model format
- GGUF
- License note
- Commercial use allowed free of charge; modified repo code must be open sourced
NobodyWho Alternatives
AakarDev AI
AakarDev AI helps teams manage AI provider access, project setup, logs, and analytics in one dashboard. BYOK support included.
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repos and verifies changes with compilers, debuggers, sanitizers, and tests.
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads in Firecracker micro-VMs with private networking and SDK, CLI, or MCP control.
hob
hob is an independent workspace for coding agents, with local control over sessions, terminals, history, routing, and follow-up work.
Ably Chat
Ably Chat is a chat API platform for custom realtime chat apps, with rooms, typing indicators, presence, reactions, message updates and usage-based pricing.
Manta AI
Manta AI is an autonomous web app testing tool that maps app behavior, catches regressions, and generates tests from a URL, no scripts or selectors needed.