shad0w watches your LLM answer the same small questions (which intent, which team, spam or not), learns the answers, and gives them back in microseconds. It only answers what it's proven to agree on; the rest goes to your LLM. Open source 🐍