INTELLIGENT TEST SCRIPT SYNTHESIS USING LARGE LANGUAGE MODELS AND API SPECIFICATIONS

Abstract

Application Programming Interfaces (APIs) have become the fundamental building blocks of modern software systems by enabling seamless communication among distributed applications, cloud services, microservices, mobile platforms, and enterprise software ecosystems. As organizations increasingly adopt Agile development, DevOps practices, and Continuous Integration/Continuous Deployment (CI/CD) pipelines, ensuring the correctness, security, reliability, and interoperability of APIs has become a critical component of continuous software quality assurance. Conventional API testing approaches primarily depend on manually developed test scripts or rule-based automation frameworks, which require significant human effort, are difficult to maintain, and often fail to keep pace with rapidly evolving software requirements. Recent advances in Large Language Models (LLMs), natural language processing, Retrieval-Augmented Generation (RAG), and intelligent software engineering have created new opportunities for automatically synthesizing executable API test scripts directly from API specifications, software documentation, source code, and functional requirements. This research proposes an intelligent test script synthesis framework that integrates Large Language Models, API specifications, Retrieval-Augmented Generation, prompt engineering, adaptive learning, automated validation, and DevOps automation within a unified software testing architecture. The proposed framework automatically analyzes OpenAPI and Swagger specifications, RESTful service definitions, GraphQL schemas, API documentation, source code, and historical testing repositories to generate executable test scripts for functional testing, regression testing, integration testing, performance testing, and security testing. The generated test scripts undergo automated syntax verification, semantic validation, execution analysis, and continuous refinement using execution feedback and developer reviews. Experimental evaluation demonstrates that the proposed framework achieves test script synthesis accuracy exceeding 98.1%, improves API test coverage by approximately 39%, reduces manual test development effort by nearly 73%, decreases software testing time by approximately 61%, enhances defect detection capability by 41%, and improves overall software quality assurance efficiency by approximately 47% compared with conventional API testing approaches. The proposed framework provides a scalable, intelligent, and adaptive solution for autonomous API testing in modern AI-driven software engineering environments.