dspy-refine

Iterative self-improvement with dspy.Refine - — wraps any module, scores each attempt with a reward function, generates feedback on failures, and retries until a quality threshold is met. Use when you want outputs to improve through self-critique, need iterative revision of drafts, or want the LM to learn from its own mistakes within a single request. Also used for self-critique and revise, iterative improvement…