Ryan-yang125/ChatLLM-Web

Private local model studio, AI chat, and agent workspace powered by WebGPU and WebLLM.

What it solves

ChatLLM Web provides a completely private, browser-native environment for interacting with Large Language Models (LLMs). It eliminates the need for external servers, API keys, or accounts by running models directly on the user's device using WebGPU, ensuring that conversations and data never leave the local machine.

How it works

The application uses WebLLM to run models in a dedicated web worker, leveraging the device's GPU via WebGPU for inference. It manages model assets (weights and WASM) through the browser's Cache API and stores conversations, preferences, and generated artifacts in IndexedDB. It includes a model studio that detects device capabilities to recommend compatible models and allows users to import custom MLC manifests via HTTPS.

Who it’s for

It is designed for users who want a private AI workspace, developers looking for a local coding assistant, and power users who want to experiment with various quantized open-source models without setting up complex local environments.

Highlights

  • Local Agent Workspace: Supports native function calling with specific models (like Hermes) to perform tasks such as searching local files and calculating arithmetic.
  • Artifact Management: Creates and updates saved artifacts with a line-diff review process before committing to local storage.
  • Extensive Model Catalog: Includes 20 curated models (chat, coding, reasoning, vision) and access to over 65 official logical models.
  • Privacy-First: No backend, no telemetry, and no API keys; all processing and storage happen locally in the browser.
  • Local Context: Allows attaching text, Markdown, JSON, and code files directly as context for the model.

Related

  • Project
    addyosmani/chatty

    A private, in-browser AI chat interface that uses WebGPU to run LLMs locally on your hardware for offline use and data privacy.

  • Project
    mlc-ai/web-llm

    A high-performance in-browser LLM inference engine that uses WebGPU to run language models locally without a server.

  • Project
    sybil-solutions/local-studio

    Local Studio is an open‑source macOS/desktop app that lets you run, manage, and chat with self‑hosted LLM back‑ends (vLLM, SGLang, llama.cpp, MLX). It combines a Bun + Hono controller (model lifecycle, OpenAI‑compatible proxy, GPU metrics) with a Next.js/Electron UI and an integrated Pi coding‑agent. The UI works locally or can point at a remote controller, and a companion mobile app (KittyLitter) can pair to reuse the same sessions. The repo provides full dev‑setup scripts, production build instructions, and a CI‑driven release pipeline.

  • Project
    ngxson/wllama

    A WebAssembly binding for llama.cpp that enables running LLMs, multimodal models, and tool calling directly in the web browser using WebGPU and Wasm.

  • Project
    sauravpanda/BrowserAI

    A JavaScript SDK that enables running production-ready LLMs and multimodal AI models directly in the browser using WebGPU acceleration for privacy and zero server costs.