NobodyWho React Native

NobodyWho is a React Native library for running large language models locally and offline on iOS and Android.

Free to use in commercial projects under the EUPL-1.2 license — no API key required. Supports text, vision, embeddings, RAG & function calling.

• Documentation — React Native & other frameworks documentation
• Starter example app — Test this library in 5 minutes
• Discord — Get help, share ideas, and connect with other developers
• GitHub Issues — Report bugs
• GitHub Discussions — Ask questions and request features

Quick Start

Install the library:

npm install react-native-nobodywho

Supported Model Format

This library uses the GGUF format — a binary format optimized for fast loading and efficient LLM inference. A wide selection of GGUF models is available on Hugging Face.

Compatibility notes:

• Most GGUF models will work, but some may fail due to formatting issues.
• For mobile devices, models under 1 GB tend to run smoothly. As a general rule, the device should have at least twice the available RAM as the model file size. Note that available RAM differs from total RAM — iOS typically reserves around 1–2 GB for the kernel and system processes, while Android overhead varies by manufacturer: roughly 2 GB on stock Android (e.g. Pixel devices), and between 2–4 GB on Samsung, Xiaomi, and Oppo devices due to additional services.

Minimum recommended specs:

• iOS: iPhone 11 or newer with at least 4 GB of RAM.
• Android: Snapdragon 855 / Adreno 640 / 6 GB RAM or better.

Model Loading

Models can be loaded from a local file path or downloaded automatically from HuggingFace:

import { Model } from "react-native-nobodywho";

// Download from HuggingFace (cached automatically)
const model = await Model.load({
  modelPath: "hf://NobodyWho/Qwen_Qwen3-0.6B-GGUF/Qwen_Qwen3-0.6B-Q4_K_M.gguf",
});

// Or load from a local file
const model = await Model.load({ modelPath: "/path/to/model.gguf" });

Downloaded models are cached on disk and reused on subsequent loads.

Chat

import { Chat } from "react-native-nobodywho";

const chat = await Chat.fromPath({
  modelPath: "hf://NobodyWho/Qwen_Qwen3-0.6B-GGUF/Qwen_Qwen3-0.6B-Q4_K_M.gguf",
  systemPrompt: "You are a helpful assistant.",
});

// Stream tokens
for await (const token of chat.ask("Is water wet?")) {
  console.log(token);
}

// Or get the full response
const response = await chat.ask("Is water wet?").completed();

See the Chat documentation for details.

Tool Calling

Give your LLM the ability to interact with the outside world by defining tools:

import { Chat, Tool } from "react-native-nobodywho";

function getWeatherForCity(city: string): string {
  return JSON.stringify({ temp: 22, condition: "sunny" });
}

const getWeather = new Tool({
  name: "get_weather",
  description: "Get the current weather for a city",
  parameters: [
    { name: "city", type: "string", description: "The city name" },
  ],
  call: getWeatherForCity,
});

const chat = await Chat.fromPath({
  modelPath: "/path/to/model.gguf",
  tools: [getWeather],
});

const response = await chat.ask("What's the weather in Paris?").completed();

See the Tool Calling documentation for more.

Sampling

The model outputs a probability distribution over possible tokens. A sampler determines how the next token is selected from that distribution. You can configure sampling to improve output quality or constrain outputs to a specific format (e.g. JSON):

import { Chat, SamplerPresets } from "react-native-nobodywho";

const chat = await Chat.fromPath({
  modelPath: "/path/to/model.gguf",
  sampler: SamplerPresets.temperature(0.2), // Lower = more deterministic
});

See the Sampling documentation for more.

Vision & Hearing

Provide image and audio information to your LLM.

To enable this, you need two model files:

• A multimodal LLM, so the LLM can consume image-tokens or/and audio-tokens
• A matching projection model, which converts images to image-tokens or/and audio to audio-tokens (usually has mmproj in the name)

Pass the projection model when loading your model, then use Prompt to compose prompts that mix text and images:

import { Chat, Prompt } from "react-native-nobodywho";

const chat = await Chat.fromPath({
  modelPath: "/path/to/vision-model.gguf",
  projectionModelPath: "/path/to/mmproj.gguf",
});

const response = await chat
  .ask(
    new Prompt([
      Prompt.Text("Tell me what you see in the image and what you hear in the audio."),
      Prompt.Image("/path/to/dog.png"),
      Prompt.Audio("/path/to/sound.mp3"),
    ]),
  )
  .completed();

You can pass multiple images/audio files and interleave text between them. If the model performs poorly, try reordering the text, audio and image parts — this can make a noticeable difference. If images consume too much context, increase contextSize or preprocess images with compression.

See the Vision & Hearing documentation for model recommendations and advanced tips.