Multiagent Microservice Architecture Generation from Textual Requirements

This repository contains a multi‑agent framework that automatically generates microservice‑oriented architectures from natural‑language software requirements using Large Language Models (Google Gemini). The pipeline is built with CrewAI and LangChain and employs two architect agents (Domain‑Driven Design and event‑driven) together with a validator agent that consolidates their outputs into a final, high‑precision decomposition. The approach is evaluated on the Spring PetClinic and Bookstore benchmark systems, replicating and extending the study “Toward Generating Microservice Architectures from Textual Requirements with LLMs”. The repository includes all source code, experiment scripts (Zero‑Shot and Few‑Shot), result datasets, statistical summaries, and detailed documentation of the experimental methodology and findings.