Wondering if there are documents for human being not for agents

First of all, thanks for the repo. It could be a gem for me to train a drafter.

However, I find it very hard to follow the code base and all the md files along the code are certainly written by an agent and not for human. Maybe it is just me but I was not able to run a toy example training run. I tried to figure it out how to launch through the code base but it has unique structures, components, and wiring up that are hard to digest. The md files are not human friendly and reading it just makes me feel more clueless.
I have to ask an agent how, for examples, starting mooncake master, starting sglang server, patching the sglang, creating vocab mapping, etc.
But I want to understand genuinely why things are wired up as is then just launching a training.

I was wondering if you guys could add documents for human…

Start by reviewing the existing Markdown files and tracing the documented toy training workflow. Cover the entry points mentioned in the issue: starting Mooncake master, starting an SGLang server, patching SGLang, and creating the vocabulary mapping. Done means a human-oriented guide explains how to run training and why these components are wired together.