Kiln AI eval workbench

Kiln is a local-first AI workbench and Python library for evals, prompt optimization, RAG, agents, synthetic data, fine-tuning, model selection, and review workflows. Why it matters: Agent and RAG workflows need repeatable checks before they are trusted. A local-first eval workbench can help teams move from ad hoc prompt testing to measured iteration.

Try this: Create a small eval set for one AI workflow, run it through Kiln, and use the results to compare two prompts or model choices.