This project provides an open-source AI assistant built with Electron, designed to enhance productivity across coding, writing, and reading tasks. It aims to offer a floating, globally accessible window for quick AI interactions, supporting various applications like VSCode, Word, and browsers, with a focus on personalized assistance and efficient content generation.

How It Works

Onepoint leverages ChatGPT technology to enable users to train the assistant for accurate content generation and reconstruction. It operates as a desktop application with a floating window, accessible via global hotkeys, minimizing desktop clutter and system resource usage. The architecture supports integration with various editing and reading applications, allowing for context-aware AI assistance.

Quick Start & Requirements

• Install via official website download.
• Development setup: git clone, cd onepoint, yarn, yarn start.
• Requires Node.js and Yarn.
• Mac platform has native support for IDE and browser content acquisition via AppleScript; Windows support is limited and planned via VBScript.

Highlighted Details

• Supports one-click code writing and refactoring across multiple IDEs.
• Acts as a translation and document writing assistant, including summarization.
• Advanced reading assistant for summarizing browser content.
• Integrates with third-party voice assistants (e.g., Xiao Ai).
• Features personalized prompts, custom character presets, and a plugin market.

Maintenance & Community

• MIT License.
• Encourages PRs and feedback; offers NFT rewards for contributors.
• Roadmap includes high availability, personalization, efficient output, reading assistance, creative play via plugins, and model training capabilities.

Licensing & Compatibility

• MIT License.
• Compatible with commercial use and closed-source linking.

Limitations & Caveats

Windows platform lacks native support for IDE code selection and browser content acquisition, relying on platform-specific scripting (AppleScript on Mac, planned VBScript for Windows). Web page crawling has a current character limit of 4000 characters.