BROWSE
Detail
Conference Paper
An LLM-driven Framework for Dynamic Infrastructure as Code Generation
Cited 0 time in 
Share 
Share
- Authors
- Junhee Lee, SungJoo Kang, In-Young Ko
- Issue Date
- 2024-12
- Citation
- International Middleware Conference (Middleware) 2024, pp.1-2
- Language
- English
- Type
- Conference Paper
- Abstract
- This paper proposes a Large Language Model (LLM)-driven framework for generation of Infrastructure as Code (IaC) in dynamic environments. While IaC simplifies infrastructure management, static templates are often inadequate. Leveraging recent LLM advancements, we design and integrate requirements refinement, retrieval-augmented generation (RAG), and code ensemble methods to improve IaC code accuracy. Preliminary results show that, despite challenges(syntax errors and hallucinations), manual adjustments enable executable code, suggesting potential for dynamic generation. CCS Concepts: • Software and its engineering → Automatic programming.
- KSP Keywords
- Dynamic Environment, Ensemble method, Infrastructure as Code, Infrastructure management, Language Model, automatic programming, code generation, preliminary results, syntax errors