meistrari/prompts-royale

Implements Monte Carlo-based matchmaking paired with ELO rating updates to efficiently rank prompt candidates across test cases, automatically generating both prompts and evaluation scenarios from a task description. The system models each prompt as a normal distribution, strategically selecting battle pairings proportional to their probability of being optimal, then updates scores through an LLM judge evaluating candidate responses. Runs entirely client-side in the browser with local storage, compatible with any LLM API.

603 stars. No commits in the last 6 months.