This project provides a well-designed, cross-platform Gemini UI, suitable for both web and desktop use. It aims to offer users a private, locally-stored, and feature-rich interface for interacting with large language models, supporting features like markdown rendering, prompt templates, and long conversation history management.

How It Works

The application is built with a focus on user experience and privacy, storing all chat data locally in the browser. It supports streaming responses and features a responsive design with dark mode and PWA capabilities. Key functionalities include markdown rendering (LaTex, mermaid, code highlighting), prompt templates for creating and sharing chat tools, and automatic compression of chat history to manage long conversations efficiently. It also offers multi-language support and can be deployed easily via Vercel or Docker.

Quick Start & Requirements

• Web Deployment: One-click deployment via Vercel using the provided button. Requires setting the CODE environment variable for access control.
• Desktop App: Downloadable clients for Linux, Windows, and macOS from the releases page.
• Docker: docker pull blacksev/gemini-next-web and docker run -d -p 3000:3000 -e OPENAI_API_KEY=sk-xxxx -e CODE=your-password blacksev/gemini-next-web.
• Development: Requires NodeJS >= 18 and Yarn. Local development involves creating a .env.local file with API keys.
• Prerequisites: OpenAI API key is required for core functionality. Azure endpoints and custom model configurations are also supported.

Highlighted Details

• Compact client size (~5MB) for desktop applications.
• Fast first-screen loading speed (~100kb).
• Support for self-hosted models via RWKV-Runner and LocalAI (llama, falcon, etc.).
• Ability to create, share, and debug chat tools using prompt templates.

Maintenance & Community

The project is actively developed, with recent updates including prompt templates, image sharing, and cross-platform desktop clients. Community interaction is facilitated via GitHub issues and a Telegram group.

Licensing & Compatibility

The project is licensed under the MIT license, which permits commercial use and linking with closed-source projects.

Limitations & Caveats

The project relies on external API keys (e.g., OpenAI) for its primary functionality. While it supports self-hosted models, setup and configuration for these may require additional effort. Automatic updates for Vercel deployments require forking the repository and enabling GitHub Actions.