Nystic-Shadow/DeepSeek-Free-API ? reverse-engineered prompt

Build me a small Python server that acts like an OpenAI compatible proxy for DeepSeek, so I can point apps at it and use /v1/chat/completions without changing much. It should read one or more DeepSeek user tokens from a text file, rotate through them if one fails, and handle the free web chat flow behind the scenes, including the anti bot challenge if needed. I want streaming responses to work, and when the model supports it, I want the reasoning text to come through too.

Please include a simple usage page so I can see requests, token counts, and errors, and make the root path look like a harmless 404 page instead of exposing the app. I also need the usual model listing and a legacy completions endpoint if that fits naturally. If you need to check current docs or behavior online, go ahead and do that.