EvaQA: An Autonomous AI-Powered QA Testing Platform with Natural Language Driven Visual Test Orchestration

Software quality assurance (QA) remains one of the most resource-intensive phases of the software development lifecycle.Traditional QA workflows rely on manual test case authoring or require engineers proficient in frameworks such as Selenium, Cypress, or Playwright-creating a significant technical barrier for non-developer stakeholders.This paper presents EvaQA, a novel desktop-based autonomous QA testing platform that eliminates code authorship from the testing process entirely.EvaQA accepts plain-English natural language descriptions of test scenarios, autonomously generates executable Playwright test scripts through a locally hosted large language model (LLM) via Ollama, and orchestrates their execution with real-time visual feedback rendered as an animated flowchart using React Flow.A proprietary DOM Context Injection Pipeline resolves the fundamental challenge of AI-generated selector hallucination by scraping live interactive elements from an embedded Chromium browser prior to each AI inference call, grounding all generated selectors in the actual runtime DOM.The platform further provides evidence capture (screenshot), structured JSON logging, automated PDF report generation, and Google Docs-style multi-role collaborative sessions powered by Liveblocks.Evaluated across 47 real-world web applications, EvaQA achieved a test script generation accuracy of 91.3%, with an average test plan generation time of 8.2 seconds on a mid-range consumer GPU.