muhammad-khalid-bin-walid/DraCo-Token-Optimizer ? reverse-engineered prompt

Build me a production ready tool that helps AI coding assistants shrink prompts and code context a lot without losing meaning. I want it to work as both a Python package and a JavaScript package, with a simple API and a command line tool. It should be able to count tokens, analyze text, and apply smart reductions to code, comments, and verbose phrasing while keeping quality high.

Please include support for the special compact text format mentioned in the docs, plus a way to turn on different optimization levels and safety checks. I also want a small dashboard or local web view for metrics, and basic benchmark reporting so I can see how much token reduction I am getting. Make it easy to run in Docker too, and if you need to look up current docs online for anything, go ahead.