I built an open-source React AI workspace that runs sandboxed Python (Wasm) and vector search entirely in the browser

srinivas191206 — Sat, 04 Jul 2026 06:35:04 +0000

Hey developers,

For the past few weeks, I’ve been working on a React-based AI workspace called Much.

I love self-hosting AI chats, but I got pretty tired of how heavy setups like LibreChat or open-webui can be. Setting up multiple Docker containers, database engines, and backend sandboxes just to run some basic code interpreter or read a PDF felt like overkill for a personal server.

So, I built Much — a lightweight, self-hostable AI workspace that runs sandboxed code execution and document searches entirely inside your browser tab using WebAssembly.

Check out the project here:
👉 GitHub Repository: github.com/srinivas191206/MUCH
👉 Live Demo: much-liard.vercel.app

How is it different?

Unlike other platforms that offload everything to heavy backend server containers, Much handles code execution and file parsing directly in your web browser tab:

In-Browser Python Sandbox: Runs Python, loads Pandas dataframes, and displays Matplotlib plots client-side using Pyodide WebAssembly. No server resources or backend sandboxes required.
Local Vector RAG: Drag-and-drop spreadsheets or PDFs and run vector-based similarity matches directly inside your browser tab.
Unified Model Hub: Hot-swap between cloud models (Gemini, Groq, OpenRouter) and local Ollama instances inside a single chat stream.
Model Context Protocol: Native MCP client support to hook models directly to local SQLite databases or file paths.
Generative UI (Artifacts): Real-time React component and Mermaid flowchart rendering in a side-by-side preview panel.

The Tech Stack

The client is built on React (Vite), Pyodide, and Lucide icons. It communicates with a lightweight Express/Node API backend that stores user data securely in a local MongoDB database.

We are in active development!

Much is currently in its early development phase. I want to promote this project and take it to the next level, and I need your help and feedback to get there!

If you like what we are building, please leave a star on the repository!

If you want to contribute features, fix bugs, or help optimize the Wasm modules, pull requests and issues are highly welcomed!

Let me know what you think in the comments below!

GitHub Repo: github.com/srinivas191206/MUCH

DEV Community: srinivas191206

I built an open-source React AI workspace that runs sandboxed Python (Wasm) and vector search entirely in the browser

How is it different?

The Tech Stack

We are in active development!