Spec2Test — From Requirements to Test Cases

Spec2Test is an AI-assisted requirements quality and testability gate designed to address problems before they reach the development stage. Software projects often begin with requirements that are ambiguous, incomplete, contradictory, or difficult to verify. These issues can remain hidden until implementation or testing, where they result in misunderstandings, rework, and wasted engineering effort. Spec2Test focuses specifically on this early requirements engineering stage rather than attempting to automate the entire software development lifecycle. Users can upload requirements in common formats including PDF, DOCX, TXT, and Markdown. The system extracts and segments the document into individual requirements and analyzes them using an LLM-assisted pipeline. Spec2Test evaluates each requirement across four quality dimensions: • Ambiguity — identifies unclear or subjective wording. • Missing Information — identifies information or constraints that may be required to make a requirement actionable. • Contradiction — identifies potentially conflicting statements or conditions. • Testability — identifies requirements that are difficult to objectively verify. A key design principle is that AI findings are not treated as absolute truth. Each finding receives a confidence-aware classification: CONFIRMED, LIKELY, POSSIBLE, or INFORMATIONAL. The system provides supporting evidence and suggested clarification rather than simply labeling a requirement as "wrong." This makes uncertainty visible and keeps the human requirements engineer in control. Once issues are identified, users can refine a requirement using the findings as context. Spec2Test generates an improved requirement along with acceptance criteria and positive, negative, and edge-case test cases. The generated artifacts remain connected through a traceability chain: Requirement → Finding → Refined Requirement → Acceptance Criteria → Test Case This creates a practical bridge between requirements engineering and testing. Instead of discovering requirement-quality problems after development has already started, teams can identify and address them earlier and immediately produce artifacts that can guide implementation and validation. The project is implemented as a lightweight hackathon MVP with a Python/FastAPI backend, SQLite database, structured LLM integration, document-processing pipeline, and web-based frontend. The architecture deliberately avoids unnecessary infrastructure and focuses on a clear, demonstrable workflow. Spec2Test does not attempt to replace human requirements engineers. Its purpose is to provide an AI-assisted quality gate that highlights potential problems, explains the evidence behind them, and helps transform clearer requirements into testable engineering artifacts.