Repolode
Submit
krisyangdev

krisyangdev/ChatLLM-Web

A browser-based local AI chat and agent workspace that runs WebLLM models on your device via WebGPU, with no backend or account.

Open-source alternative to:ChatGPT

View on GitHubWebsite
Stars
640
Forks
48
Open issues
4
Last commit
3 months ago

What is it?

ChatLLM Web is a static web app for chatting with large language models that run entirely in the browser. It uses WebLLM in a dedicated Web Worker to execute models on the local GPU through WebGPU. The app inspects the device's WebGPU support, memory and storage, then recommends a model and shows download, load, cache and compatibility status.

Beyond chat, it includes an Agent mode where models with native function calling (such as Hermes 2 Pro and Hermes 3) can use sandboxed local tools to read and search attached files, do arithmetic, and create or update Markdown, code, JSON or text artifacts. Every artifact write is shown as a line diff for approval before it is saved to IndexedDB. Conversations, files and artifacts stay in the browser, and the app can be installed as a PWA to reuse cached models offline after the first load.

The project is deployed on Cloudflare Pages and has no application backend, accounts, API keys or telemetry. Model weights are fetched directly from Hugging Face or WebLLM library URLs. It supports English and Simplified Chinese interfaces and light and dark themes.

Key features

  • 01Runs WebLLM models on the local GPU through WebGPU in a dedicated worker.
  • 02Offers a curated set of 20 chat, coding, reasoning, vision and tool-capable models, plus an advanced catalog.
  • 03Agent mode with sandboxed local tools for reading selected files, searching, and calculating.
  • 04Requires diff approval before any artifact is written to IndexedDB.
  • 05Attaches TXT, Markdown, JSON and common code files as direct context.
  • 06Installs as a PWA and reuses cached models after the first load.

Use cases

  • Chatting with an LLM privately without sending prompts to a cloud API.
  • Trying small open models on a laptop with a supported GPU and browser.
  • Asking questions about local code or text files without uploading them.
  • Experimenting with local tool-calling agents that work only on selected files.

Quick start

  1. 1

    Install dependencies

    npm install
  2. 2

    Start the development server

    npm run dev

Things to know

  • Requires a current WebGPU-capable browser and a device with suitable GPU memory.
  • Three Agent models are marked experimental, and the Advanced catalog is for power users.

In collections

Open-source AI agent frameworks & appsRun LLMs locally

Topics

#local-ai#webgpu#webllm#llm#browser-ai#pwa#react#ai-agents#cloudflare-pages#llama

FAQ

What is ChatLLM-Web?
A browser-based local AI chat and agent workspace that runs WebLLM models on your device via WebGPU, with no backend or account. ChatLLM Web is a static web app for chatting with large language models that run entirely in the browser. It uses WebLLM in a dedicated Web Worker to execute models on the local GPU through WebGPU. The app inspects the device's WebGPU support, memory and storage, then recommends a model and shows download, load, cache and compatibility status.
Is ChatLLM-Web free to use? What license does it have?
ChatLLM-Web is open source under the MIT license. Check the license terms before using it in commercial products.
Is ChatLLM-Web still maintained?
Updated within 6 months. The last commit to ChatLLM-Web was on Aug 12, 2026.
What can I use ChatLLM-Web for?
Chatting with an LLM privately without sending prompts to a cloud API. Trying small open models on a laptop with a supported GPU and browser. Asking questions about local code or text files without uploading them. Experimenting with local tool-calling agents that work only on selected files.
How do I install ChatLLM-Web?
Install dependencies: npm install Start the development server: npm run dev
Where can I get ChatLLM-Web?
The source code and installation instructions are on GitHub: https://github.com/krisyangdev/ChatLLM-Web

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects