Clean up Claude 5's token vomit with a separate LLM

A GitHub project called "Vomit" uses a local LLM to convert Claude's verbose token output into clearer English, running entirely offline without external dependencies or telemetry. The tool pipes Claude's responses through another language model to improve readability, though it comes with limitations including potential hallucinations and slowness. Installation requires setting up a local LLM like Llama.app or Ollama, then using the provided commands to translate Claude's outputs for specified sessions.