lab-i40/Capability-Algorithms: v2.0.0 - Full TC Matrix, Redesigned Algorithm, and Experimental Results

What's new in v2.0.0 This release closes the full experimental loop: a redesigned matching pipeline, a complete 16-case test matrix, and the recorded outcomes that support the paper's evaluation section. Algorithm Rewrote the three-step matching pipeline (semantic matching → requirement matching → aggregation) with new prompts for each step Added Chain-of-Thought guidance to both Step 1 and Step 2 prompts Enforced fenced JSON responses for reliable structured output parsing Defined the ground-truth schema for test case validation Dataset Completed all 16 test cases across the 2×8 matrix (four abstraction dimensions × two formalism levels, positive and negative pair per cell) Cases cover: screwing (formal/informal), joining, housing-cover screwing, robot/manual resource specificity, and cross-hierarchy process matching Evaluation Parametrized the test suite to run all TC cases in a single pass Added per-case diagnostic report output (semantic_matching.md, requirement_matching.md) for qualitative analysis Recorded aggregate result CSVs for the full TC matrix Security & configuration API key now read from the OPENAI_API_KEY environment variable instead of a literal; added .env template and documentation Docs Rewrote README from the paper context, covering setup, test execution workflow, and the TC evaluation matrix structure Documented report artifacts and their relationship to the paper's results section