Test4You: An LLM-based Framework for Enhancing the Software Testing Process

Test4You: An LLM-based Framework for Enhancing the Software Testing Process Vincenzo Suriani Dept. of Computer, Control and Manag. Eng. University of Rome La Sapienza Rome, Italy [email protected] Monica Sileo Dept. of Engineering University of Basilicata Potenza, Italy [email protected] Domenico D. Bloisi Dept. of Int. Humanities and Social Sciences International University of Rome UNINT Rome, Italy [email protected] Abstract—Ensuring the robustness and usability of web applications requires testing methodologies that go beyond traditional functional and performance evaluations. This paper presents Test4You, an LLM-based testing framework that integrates automated stress testing, accessibility evaluation, and hybrid testing workflows with real and synthetic agents. Test4You enables comprehensive analysis of web applications by combining advanced visualization tools, moderated and unmoderated execution modes, and an intelligent logging system. The proposed framework provides interactive diagnostics through tree-based and pagewise visualizations, pushes the limits of automated checks on accessibility tests, and leverages synthetic agents guided by LLMs for autonomous execution. This innovative design offers both scalability and human-centered adaptability, addressing key challenges in next-generation web testing. Index Terms—Automated Testing, Software Testing, Large Language Models. I. INTRODUCTION Test4You is a software testing framework based on Large Language Models (LLMs) developed in the context of the Tech4You project (https://ecs-tech4you.it/) funded by the European Union under the Next Generation EU plan. Tech4You is a research and innovation (R&I) program aimed at creating an innovation ecosystem between four public universities, six research centers, nine private actors (affiliated companies), three public bodies (natural parks and environmental agencies), one NGO, and two regional governments. Inspired by the Horizon Europe mission ”Adaptation to Climate Change,” the project aims to respond to the climate crisis by improving community resilience and reducing regional economic disparities. II. TESTING FRAMEWORK OVERVIEW The Test4You framework is part of the testing process for software solutions developed in Tech4You. It integrates a suite of tools and features designed for automated and supervised web environment testing. Test4You is structured around two core components (see Fig. 1): 1) a Stress Testing Suite, designed to evaluate application performance under high-intensity traffic and diverse user interactions, and 2) an AI-powered Testing Suite, embedded in a playground This study was partially funded by the Next Generation EU—Italian National Recovery and Resilience Plan (NRRP), Mission 4, Component 2, Investment 1.5, project Tech4You–Technologies for climate change adaptation and quality of life improvement, n. ECS0000009 – CUP C43C22000400006. Fig. 1. Test4You includes an automatic stress tester and a playground interactive environment with safety measures. environment for automatic and human testers. It enables automated, moderated, and unmoderated workflows with real and synthetic agents. This dual structure allows the framework to serve multiple goals simultaneously: ensure robust performance, validate accessibility, and provide transparent reporting for collaborative debugging and iterative development cycles. A playground environment unifies these modules, allowing for a seamless transition between stress testing, accessibility evaluation, and agent-based workflows. III. STRESS TESTING CAPABILITIES The Stress Testing Suite is meant to evaluate reliability, responsiveness, and stability under high load. After supplying the target URL in the playground, testers can trigger a stress session from the top control bar; the front-end hands off to a webdriver-based back-end that systematically explores page functionality at scale, emulating high-intensity traffic patterns and diverse interaction paths. Beyond raw load generation, the suite bundles featurecomplete checks that more closely mirror real usage: (i) functional testing of critical flows (clicks, form submissions, navigation), (ii) cross-browser consistency assessment, (iii) data-driven scenarios to probe variations in inputs and edge cases, and (iv) UI checks for layout integrity, responsiveness, and interactive affordances under stress. 2025 I-RIM Conference October 17-19, Rome, Italy ISBN: 9788894580570 10.5281/zenodo.17629832 197 Fig. 2. Agent’s path visualization after a stress test. IV. REPORTING AND ANALYSIS Test4You emphasizes explainable test outcomes through dual complementary visualization modes. Agent-Path (Tree/Dictionary) Visualization. Results are organized by the exploratory trajectory of the agent. Each interaction (e.g., a button press that opens a new page) expands a child node, forming an interactive tree of visited states (see Fig. 2). The nodes expose an on-demand, graphical report of the elements tested at that depth, and the viewer supports pan/zoom, drag operations, and full-screen mode. Page-Wise Visualization. Complementing path-centric views, the page-wise mode aggregates results by visited pages. Each column represents a page; rows enumerate tested components with status chips that convey three levels at a glance: success (green), attention (grey), and failure (red). Together, the above described visualization modes bridge developer and QA workflows: the agent-path tree explains how a failure was reached, while the page-wise grid summarizes where issues cluster across the application surface. V. MODERATED AND UNMODERATED TESTING Beyond load and functional checks, Test4You introduces a playground for controlled human-in-the-loop sessions, as well as autonomous agent trials, represented by multimodal LLMs including the well-known Grok and GPT models. Moderated Testing. In moderated mode, either real users or synthetic agents operate under the supervision of a facilitator. The system supports JSON-defined task guidelines, step-bystep instructions, and real-time monitoring panels. Superusers can dynamically adjust descriptions, validate actions against ground truth, and record qualitative insights such as user satisfaction or workflow friction points. This hybrid approach captures nuances that purely automated tests miss. Unmoderated Testing. For scale and repeatability, unmoderated sessions delegate tasks to autonomous synthetic agents. These agents, powered by Visual Language Models, parse the interface and plan interactions such as clicking, form entry, or menu navigation. Each agent embodies a distinct behavioral profile, allowing for diversity in coverage. Logs and execution panels track action sequences and compare them against loaded ground-truth plans, producing per-step and overall success rates. This design balances autonomy with transparency, since deviations and errors are automatically flagged for further review. VI. ACCESSIBILITY TESTING Accessibility is a central pillar of Test4You, ensuring compliance with the WCAG guidelines. The suite provides Keyboard Usability, Contrast Ratio Analysis, Screen Reader Support, Logical Navigation & Responsive Design. All of these features transform accessibility validation from an afterthought into an integrated testing dimension. VII. TASK METRICS AND LOGGING Quantitative evaluation underpins the robustness of the framework. Metrics include: •Element Accuracy: measuring correct targeting of interactive components. •Operation F1 Score: balancing precision and recall for atomic actions (e.g., selecting an option). •Step & Task Success Rates: aggregated completion statistics for workflows. These indicators are complemented by a real-time logging system that records clicks, keystrokes, mouse movements, and navigation events. Logs are structured, exportable, and filterable by event type, creating a transparent trace of execution. This supports both debugging and compliance auditing. VIII. CONCLUSIONS The validation activities conducted for the Test4You framework represent a contribution to creating a robust and versatile testing environment for web applications. By integrating a collection of testing modalities, such as stress testing, accessibility evaluations, moderated and unmoderated workflows, and advanced reporting mechanisms, the platform aims to improve the reliability, usability, and inclusivity of digital systems. The framework addresses diverse user and application needs: 1) inclusion of advanced visualization techniques, such as tree-based and page-wise formats, enables a more intuitive analysis of testing results, facilitating efficient issue identification and collaborative problem solving; 2) the dual focus on synthetic and real-user testing workflows ensures the platform’s scalability while maintaining alignment with realworld usage scenarios. 198