7as0nch/mimo2codex — reverse-engineered prompt

Build me a local desktop and CLI friendly proxy app that lets the newest OpenAI Codex CLI and Codex desktop app use other AI models without downgrading Codex.

It should run on my own computer at localhost, translate Codex’s Responses API requests into Chat Completions style requests when needed, and route requests by the model name. Please include built in setup for Xiaomi MiMo and DeepSeek, plus a way to add other OpenAI compatible providers like Qwen, GLM, Kimi, Ollama, LM Studio, vLLM, or OpenAI itself through config instead of code changes.

I want a simple admin web page where I can see providers, model aliases, logs, token usage, and settings. It should store data locally, support API keys safely, and work from npm, Docker, and a Windows or macOS desktop app that can run in the background and start on boot.

Please make it practical for non technical users, with clear install steps, troubleshooting, and Codex config generation.