| { |
| "name": "python_engineering_signals", |
| "description": "Classify fixed-size Python code blocks for whether they exhibit common OSS-style engineering practices (documentation, typing, structured logging, tests, defensive errors).", |
| "language": "python", |
| "conventions": [ |
| { |
| "name": "has_type_hints", |
| "description": "Function or method definitions use PEP 484 annotations (parameters and/or return), or module-level variables use explicit type annotations." |
| }, |
| { |
| "name": "has_docstring_or_comment_block", |
| "description": "The block includes a docstring (module, class, or function) or a substantial comment block explaining intent (not only # noqa or trivial one-liners)." |
| }, |
| { |
| "name": "uses_logging_not_print", |
| "description": "Uses the logging module (e.g. logger.info, getLogger) for diagnostic output rather than bare print for operational messages." |
| }, |
| { |
| "name": "test_or_pytest_artifacts", |
| "description": "Test-oriented code: pytest/unittest imports, @pytest.fixture, test_* functions, assert used in a test-like context, or mock/patch from unittest.mock." |
| }, |
| { |
| "name": "defensive_error_handling", |
| "description": "Explicit failure paths: try/except with handling, raise ValueError/TypeError with a message, assert with a message, or guard clauses that raise early." |
| } |
| ], |
| "block_lines": 48, |
| "test_repos": 3, |
| "train_repos": 8, |
| "val_repos": 2, |
| "target_f1": 0.78, |
| "max_repair_rounds": 3, |
| "gliner2_base_model": "fastino/gliner2-base-v1", |
| "gliner2_hf_repo": "gilpasternak-cm/gliner2_classifier" |
| } |
