Can LLMs Detect Microservice Infrastructure Patterns Across Languages?

arXiv·medium signal

Evaluates whether LLMs can detect architectural patterns (service mesh, circuit breaker, API gateway, etc.) in microservice infrastructure artifacts written in multiple languages — a task current tools struggle with because they only support single-language detection. LLMs trained on diverse software artifacts may overcome this limitation. The study identifies factors influencing detection accuracy and failure modes, providing practical guidance for teams using LLMs as architecture analysis tools.