CTLab-ITMO/CoolPrompt

Implements multiple optimization algorithms (HyPE, ReflectivePrompt, DistillPrompt) that iteratively refine prompts through LLM-based feedback and evaluation metrics. LLM-agnostic architecture supports any Langchain-compatible model while generating synthetic evaluation data when datasets are unavailable, and automatically detects task types for scenarios without explicit specifications.

178 stars and 84 monthly downloads. Available on PyPI.