🧠 Local LLM Chatbot (WebLLM + React)

This is a chatbot powered by @mlc-ai/web-llm that runs completely in the browser — no internet or server required after setup!

🚀 Features

• Uses RedPajama 3B model
• Fully runs in-browser using WebGPU (no API key needed)
• Works on modern desktop browsers (Chrome, Edge)
• Lightweight UI built with React + Tailwind CSS

Notes

• WebGPU support is required (only works on modern Chromium-based browsers like Chrome/Edge).
• Works best on desktop/laptops with GPU support.
• May not work on low-end or mobile devices due to limited GPU power and WebGPU support.

---

🚨 Important Note About Vercel Hosting While the application is fully functional when hosted on Vercel, please note the following:

The WebLLM model is downloaded entirely on the client-side, which means initial loading times can be long (especially on slower networks or devices).

You might see a delay before the model starts responding — this is expected due to model weight downloading and initialization in-browser.

✅ You can still explore the full UI and user experience via the hosted link.

⚠️ Recommendation: For a smooth and responsive experience, we highly recommend cloning the repository locally and running the app using:

npm install npm run dev

This way, the model will only be downloaded once and then reused, resulting in faster interactions and testing.

---

📦 Installation & Setup

1. Clone the Repo

git clone https://github.com/your-username/web-llm-chat.git
cd web-llm-chat

### 2. Install dependencies

npm install
Start the development server

npm run dev
Model Download & Loading

As soon as you run the dev server, the app will automatically start downloading the selected LLM model (e.g., RedPajama-INCITE-Chat-3B).

You’ll see progress logs in the console like Model loading progress: XX%.

Once the loader disappears from the UI, the chatbot is ready for use ✅