FiePaw/PAF-ModelQwen — reverse-engineered prompt

Build me a Python tool that uses the Qwen web chat in a real browser so I can send prompts, keep sessions alive with saved cookies, and avoid logging in every time. I want it to handle multiple Qwen accounts and automatically switch if one hits a rate limit, expires, or runs out of quota. It should support think modes like auto, deep thinking, and fast, and let me start a new chat or continue the last one.

I also want it to be practical for real use, so please make it async, keep a small pool of browsers ready in the background, save logs and responses, and optionally save code blocks from replies. It should support batch prompts, attachments pasted into the chat, and the built in Qwen actions for web search, image generation, and video generation if available through the UI.

If possible, include a simple local worker plus a VPS bridge so I can expose it as an API with an OpenAI style chat endpoint. Look up current docs online if you need to.