A deterministic LLM agent makes a copy of its initial run, saves an artifact of what and how it did it, so that it can be replicated by the instructional deterministic machine GitHub: https://github.com/apervmm/llm-use-automation
