Using a Feedback Loop for LLM-based Infrastructure as Code Generation

Code generation with Large Language Models (LLMs) has helpedto increase software developer productivity in coding tasks, but hasyet to have significant impact on the tasks of software developersthat surround this code. In particular, the challenge of infrastructuremanagement remains an open question. We investigate the abilityof an LLM agent to construct infrastructure using the Infrastructureas Code (IaC) paradigm. We particularly investigate the use of afeedback loop that returns errors and warnings on the generatedIaC to allow the LLM agent to improve the code. We find that, foreach iteration of the loop, its effectiveness decreases exponentiallyuntil it plateaus at a certain point and becomes ineffective.