This project provides a Raycast extension for free access to various large language models, including GPT-4, Llama-3, and Claude, with support for custom APIs. It targets Raycast users on macOS seeking a cost-effective and flexible alternative to paid AI solutions, offering features like streaming responses, web search, and file uploads.

How It Works

The extension leverages a network of third-party providers and models, allowing users to select and switch between different AI services. It integrates with the Raycast application, enabling AI interactions directly from the command palette. Key features include real-time streaming, web search capabilities via DuckDuckGo, file uploads (text, image, video, audio), image generation, and custom AI command creation.

Quick Start & Requirements

• Install the Raycast app (macOS only, Windows support in progress).
• Install Node.js (v20.18.1 recommended).
• Installation from source: clone repo, run npm ci --production, optionally pip3 install -r requirements.txt, then npm run dev.
• Official Docs: Homepage

Highlighted Details

• Supports 18 commands, including AI chat, web search, and file uploads.
• Integrates with numerous providers like Nexra, DeepInfra, Blackbox, and Google Gemini.
• Offers custom API support for OpenAI-compatible endpoints.
• Features a "Code Interpreter (BETA)" for select DeepInfra models.

Maintenance & Community

• Updates are managed via an in-app "Check for Updates" command or manual git pull, npm ci --production, npm run dev.
• Community support is available via GitHub issues.
• Discord link: Discord

Licensing & Compatibility

• License: GPLv3.
• Derived from Raycast Gemini by Evan Zhou.
• Uses third-party libraries g4f-image and gemini-ai-sdk.
• Code inspired by the original gpt4free Python project.

Limitations & Caveats

The extension is not available on the official Raycast store due to provider verification policies and update speed concerns. Windows support is in progress. The UI is limited by Raycast's developer API. Some features, like code interpretation, are in beta and may have limited model support.