youssefvdel/qwen-gate — reverse-engineered prompt

Build me a local TypeScript app called Qwen Gate that lets me use chat.qwen.ai through an API that works like OpenAI. I want to run it on my computer, add several Qwen accounts in a simple web dashboard, and have requests automatically rotate between them so cooldowns and failed accounts are handled without me thinking about it.

The API should support /v1/chat/completions and /v1/models, normal chat responses, streaming responses, and tool calling in the OpenAI style as much as possible. It should clean up weird model output like thinking tags, repeated echoes, or tool call junk before sending results back.

Please include a dashboard where I can see status, accounts, request logs, network activity, and settings. Use browser automation to log in and keep Qwen sessions alive. Add a config file for port, API key, browser choice, logging, retries, and cleanup options. Make it easy to install and run on Windows, Mac, and Linux.