A ChatGPT client that lives entirely in your browser

What it does Light-GPT is a pure-frontend chat interface built with Next.js that talks directly to OpenAI’s GPT-3.5-Turbo API from your browser. You supply your own API key, and the app streams replies with a typewriter effect while storing every conversation in IndexedDB. It also handles text-to-image generation and exports chats as images or PDFs, all without touching a backend server.

The interesting bit By making API calls straight from the client, the project eliminates backend costs and server-side data leakage entirely. The README is unusually frank about the downsides: because requests originate in your browser, you need direct, unfiltered internet access to OpenAI, and the authors warn of “abnormal risk” if your connection triggers OpenAI’s risk controls.

Key highlights

• Zero-backend architecture: static deployment with no server-side code.
• Strictly local storage: your API key and all chat history remain in the browser.
• Export conversations as images or PDFs; includes syntax highlighting and one-click code copying.
• Supports text-to-image generation alongside standard chat.
• Custom avatars and responsive design for both desktop and mobile.

Caveats

• Requires direct browser access to OpenAI; VPNs or regional blocks will prevent it from working.
• The README explicitly warns users not to enter an API key unless they can already browse OpenAI’s site normally.
• Still locked to GPT-3.5-Turbo with no evident path to newer models.

Verdict Ideal for privacy-minded developers who want a personal ChatGPT UI without maintaining a backend. Not suitable for teams, users behind restrictive networks, or anyone needing model flexibility beyond GPT-3.5.