mudler/LocalAI ? reverse-engineered prompt

I want to build a self-hosted AI engine that I can run completely locally on my own hardware. It should be an all-in-one solution for running various open-source models, functioning like a private, personal alternative to services like OpenAI, Anthropic, or Midjourney.

The system needs to handle text generation, image generation, audio transcription, and text-to-speech. A critical feature is that it should run on any hardware I have—whether it's just a CPU or a machine with an NVIDIA, AMD, or Apple Silicon GPU.

Most importantly, please make its API a drop-in replacement for the OpenAI API. This will let me use existing client libraries and tools without changing my code. It should also include a simple web UI to manage the models, view stats, and maybe handle some basic user authentication. The entire application should be packaged to run inside a Docker container for easy deployment.