QwenLM/qwen-code ? reverse-engineered prompt

I want to build an open-source AI agent that runs right in the terminal, a sort of coding copilot for the command line. When a user runs the tool, it should open up an interactive chat session where they can ask for help with their code.

The key feature is letting users connect to their favorite AI models. It should support using API keys for major providers like OpenAI, Anthropic, and Google's Gemini. I also want to include a free-tier option where users can log in with an account, maybe through a browser, to get a limited number of daily requests without needing an API key.

Inside the chat session, there should be simple slash commands like /auth to manage logins and /model to switch between different AIs. It should be easy for users to set up their models and keys in a configuration file. Let's use Node.js and TypeScript for this, and make sure the final tool is easy to install on Mac, Linux, and Windows with a simple one-line script.