README.md

LLM.js

Run Large-Language Models (LLMs) 🚀 directly in your browser!

Example projects🌐✨: Live Demo

Learn More: Documentation

Models Supported:

• TinyLLaMA Series - 1,2,3🦙
• GPT-2
• Tiny Mistral Series
• Tiny StarCoder Py
• Qwen Models
• TinySolar
• Pythia
• Mamba and much more✨

Features

• Run inference directly on browser (even on smartphones) with power of WebAssembly
• Guidance: Structure responses with CFG Grammar and JSON schema
• Developed in pure JavaScript
• Web Worker to perform background tasks (model downloading/inference)
• Model Caching support
• Pre-built packages to directly plug-and-play into your web apps.

Installation

Download and extract the latest release of the llm.js package to your web application📦💻.

Quick Start

// Import LLM app
import {LLM} from "llm.js/llm.js";

// State variable to track model load status
var model_loaded = false;

// Initial Prompt
var initial_prompt = "def fibonacci(n):"

// Callback functions
const on_loaded = () => { 
    model_loaded = true; 
}
const write_result = (text) => { document.getElementById('result').innerText += text + "\n" }
const run_complete = () => {}

// Configure LLM app
const app = new LLM(
     // Type of Model
    'GGUF_CPU',

    // Model URL
    'https://huggingface.co/RichardErkhov/bigcode_-_tiny_starcoder_py-gguf/resolve/main/tiny_starcoder_py.Q8_0.gguf',

    // Model Load callback function
    on_loaded,          

    // Model Result callback function
    write_result,       

     // On Model completion callback function
    run_complete       
);

// Download & Load Model GGML bin file
app.load_worker();

// Trigger model once its loaded
const checkInterval = setInterval(timer, 5000);

function timer() {
    if(model_loaded){
            app.run({
            prompt: initial_prompt,
            top_k: 1
        });
        clearInterval(checkInterval);
    } else{
        console.log('Waiting...')
    }
}