BrowserAI

BrowserAI 🚀

Run Production-Ready LLMs Directly in Your Browser

Live Demo • Documentation • Discord Community

🎉 Featured Project: Check out Browseragent.dev - A no-code AI Agent builder powered by BrowserAI with unlimited executions! Build your own AI agents in minutes.

🌟 Live Demos

Demo	Description	Try It
Chat	Multi-model chat interface	chat.browserai.dev
Voice Chat	Full-featured with speech recognition & TTS	voice-demo.browserai.dev
Text-to-Speech	Powered by Kokoro 82M	tts-demo.browserai.dev

⚡ Key Features

• 🔒 100% Private: All processing happens locally in your browser
• 🚀 WebGPU Accelerated: Near-native performance
• 💰 Zero Server Costs: No complex infrastructure needed
• 🌐 Offline Capable: Works without internet after initial download
• 🎯 Developer Friendly: Simple sdk with multiple engine support
• 📦 Production Ready: Pre-optimized popular models

🎯 Perfect For

• Web developers building AI-powered applications
• Companies needing privacy-conscious AI solutions
• Researchers experimenting with browser-based AI
• Hobbyists exploring AI without infrastructure overhead
• No-code platform builders creating AI-powered tools

✨ Features

• 🎯 Run AI models directly in the browser - no server required!
• ⚡ WebGPU acceleration for blazing fast inference
• 🔄 Seamless switching between MLC, Transformers, Flare, and Demucs engines
• 📦 Pre-configured popular models ready to use
• 🛠️ Easy-to-use API for text generation and more
• 🔧 Web Worker support for non-blocking UI performance
• 📊 Structured output generation with JSON schemas
• 🎙️ Speech recognition and text-to-speech capabilities
• 🎵 Audio source separation (Demucs) - isolate vocals, drums, bass, and other stems
• 💾 Built-in database support for storing conversations and embeddings

🚀 Quick Start

npm install @browserai/browserai

OR

yarn add @browserai/browserai

Basic Usage

import { BrowserAI } from '@browserai/browserai';

const browserAI = new BrowserAI();

// Load model with progress tracking
await browserAI.loadModel('llama-3.2-1b-instruct', {
  quantization: 'q4f16_1',
  onProgress: (progress) => console.log('Loading:', progress.progress + '%')
});

// Generate text
const response = await browserAI.generateText('Hello, how are you?');
console.log(response.choices[0].message.content);

📚 Examples

Text Generation with Options

const response = await browserAI.generateText('Write a short poem about coding', {
  temperature: 0.8,
  max_tokens: 100,
  system_prompt: "You are a creative poet specialized in technology themes."
});

Chat with System Prompt

const ai = new BrowserAI();
await ai.loadModel('gemma-2b-it');

const response = await ai.generateText([
  { role: 'system', content: 'You are a helpful assistant.' },
  { role: 'user', content: 'What is WebGPU?' }
]);

Chat with System Prompt

const response = await browserAI.generateText('List 3 colors', {
  json_schema: {
    type: "object",
    properties: {
      colors: {
        type: "array",
        items: {
          type: "object",
          properties: {
            name: { type: "string" },
            hex: { type: "string" }
          }
        }
      }
    }
  },
  response_format: { type: "json_object" }
});

Speech Recognition

const browserAI = new BrowserAI();
await browserAI.loadModel('whisper-tiny-en');

// Using the built-in recorder
await browserAI.startRecording();
const audioBlob = await browserAI.stopRecording();
const transcription = await browserAI.transcribeAudio(audioBlob, {
  return_timestamps: true,
  language: 'en'
});

Text-to-Speech

const ai = new BrowserAI();
await ai.loadModel('kokoro-tts');
const audioBuffer = await browserAI.textToSpeech('Hello, how are you today?', {
  voice: 'af_bella',
  speed: 1.0
});// Play the audio using Web Audio API
const audioContext = new AudioContext();
const source = audioContext.createBufferSource();
audioContext.decodeAudioData(audioBuffer, (buffer) => {
  source.buffer = buffer;
  source.connect(audioContext.destination);
  source.start(0);
});

Audio Source Separation (Demucs)

import { DemucsEngine } from '@browserai/browserai/demucs';

const engine = new DemucsEngine();
await engine.loadModel({ /* htdemucs config */ });

// Separate an AudioBuffer into stems
const result = await engine.separate(audioBuffer, {
  shifts: 1,        // Time-shift augmentation passes (higher = better quality, slower)
  overlap: 0.25,    // Segment overlap ratio
});

// result.sources contains: drums, bass, other, vocals (each as AudioBuffer)
const vocals = result.sources['vocals'];

Flare Engine (GGUF Models via WASM)

const ai = new BrowserAI();

// Load a GGUF model via the Flare engine
await ai.loadModel('llama-3.2-1b-flare');

// Generate text — same API as MLC/Transformers
const response = await ai.generateText('Explain quantum computing briefly');

// Optional: Load a LoRA adapter
await ai.loadAdapter({ url: 'https://example.com/adapter.safetensors' });

🔧 Supported Models

More models will be added soon. Request a model by creating an issue.

MLC Models

• Llama-3.2-1b-Instruct
• Llama-3.2-3b-Instruct
• Hermes-Llama-3.2-3b
• SmolLM2-135M-Instruct
• SmolLM2-360M-Instruct
• SmolLM2-1.7B-Instruct
• Qwen-0.5B-Instruct
• Gemma-2B-IT
• TinyLlama-1.1B-Chat-v0.4
• Phi-3.5-mini-instruct
• Qwen3-0.6B
• Qwen3-1.7B
• Qwen3-4B
• Qwen3-8B
• Qwen2.5-1.5B-Instruct
• DeepSeek-R1-Distill-Qwen-7B
• DeepSeek-R1-Distill-Llama-8B
• Snowflake-Arctic-Embed-M-B32
• Snowflake-Arctic-Embed-S-B32
• Snowflake-Arctic-Embed-M-B4
• Snowflake-Arctic-Embed-S-B4

Transformers Models

• Llama-3.2-1b-Instruct
• Whisper-tiny-en (Speech Recognition)
• Whisper-base-all (Speech Recognition)
• Whisper-small-all (Speech Recognition)
• Kokoro-TTS (Text-to-Speech)

Flare Models (GGUF via WASM)

• SmolLM2-135M-Instruct (Q8_0, Q4_K_M)
• SmolLM2-360M-Instruct (Q8_0)
• Qwen2.5-0.5B-Instruct (Q4_K_M)
• Llama-3.2-1B-Instruct (Q8_0, Q4_K_M)

Demucs Models (Audio Source Separation)

• HTDemucs (4-stem: drums, bass, other, vocals)

🗺️ Enhanced Roadmap

Phase 1: Foundation

• 🎯 Simplified model initialization
• 📊 Basic monitoring and metrics
• 🔍 Simple RAG implementation
• 🛠️ Developer tools integration

Phase 2: Advanced Features

• 📚 Enhanced RAG capabilities
  • Hybrid search
  • Auto-chunking
  • Source tracking
• 📊 Advanced observability
  • Performance dashboards
  • Memory profiling
  • Error tracking

Phase 3: Enterprise Features

• 🔐 Security features
• 📈 Advanced analytics
• 🤝 Multi-model orchestration

🤝 Contributing

We welcome contributions! Feel free to:

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• MLC AI for their incredible mode compilation library and support for webgpu runtime and xgrammar
• Hugging Face and Xenova for their Transformers.js library, licensed under Apache License 2.0. The original code has been modified to work in a browser environment and converted to TypeScript.
• Facebook Research for Demucs, the state-of-the-art music source separation model.
• Aspect for the Flare WASM inference engine enabling GGUF model support in the browser.
• All our contributors and supporters!

---

🚀 Requirements

• Modern browser with WebGPU support (Chrome 113+, Edge 113+, or equivalent)
• For models with shader-f16 requirement, hardware must support 16-bit floating point operations