Reproduce “LLM4FaaS: No-Code Application Development using LLMs and FaaS” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Minghe Wang, Tobias Pfandzelter, Trever Schirmer, David Bermbach

LLM4FaaS enables non-technical users to develop and deploy custom applications using natural language by integrating Large Language Models (LLMs) with Function-as-a-Service (FaaS) platforms. The system achieved a 71.47% semantic pass rate, demonstrating an effective bridge between LLM code generation and automated, operational deployment.

Key insights

  • FaaS abstraction is critical for enabling LLMs to focus on generating functionally correct code, significantly improving the semantic pass rate of applications for non-technical users.
  • A dedicated orchestration layer (LLM-FaaS Bridge) is essential for seamlessly connecting LLM code generation with automated FaaS deployment, abstracting away operational complexities.
  • While LLM generation accounts for the majority of the latency, the end-to-end process from natural language prompt to deployed FaaS function remains efficient, averaging around 23 seconds.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.