decodingai-magazine/llm-twin-course — reverse-engineered prompt

Build me an LLM twin project that can learn my writing style from my online content and let me chat with it or generate new posts in my voice.

I want it to collect data from places like Medium, Substack, and GitHub, clean it, store it, turn it into searchable chunks, and use RAG so the answers are based on my actual writing. Please make it a real Python project, not just notebooks, with separate parts for data collection, feature processing, model training, and inference.

Include a way to fine tune a model with LoRA or QLoRA, track experiments, save the best model, and serve it through an API. Add a simple Gradio interface where I can type a prompt and get content that sounds like me.

Use Docker for local setup where it makes sense, include clear install and usage steps, environment variable examples, and keep the code organized so I can follow it as a learning project. Look up current docs online if needed.