Detailed atom trace 2

[ColtonPayne]

Summary

• Extends the rule trace tuple from 5 fields (timestep, fp_op, component, label, bound) to 9 fields, adding consistent (bool), triggered_by (Fact/Rule/IPL), meta_name (fact/rule name), and inconsistency_message (descriptive explanation)
• Adds detailed inconsistency messages that explain why a conflict occurred — distinguishing IPL-based conflicts (e.g., sick vs healthy) from same-predicate conflicting bounds, and including the specific groundings and bounds involved
• Adds new columns to rule trace output: Consistent, Triggered By, and Inconsistency Message in the CSV/DataFrame output, making inconsistency resolution fully explainable without parsing free-text strings
• Preserves ground atom clauses in inconsistency trace entries when the conflict was triggered by a rule (passes qn/qe through instead of empty lists)
• Adds inconsistency_example.py demonstrating node and edge inconsistencies (IPL-based and same-predicate) with the new trace output
• Adds functional test (test_inconsistency_trace.py) parameterized across regular, fp, and parallel modes validating the new trace columns
• Updates unit tests to work with the new 9-tuple format, using [:5] slicing for backward-compatible assertions and adding new assertions for metadata fields
• Fixes filter.py unpacking to handle the expanded tuple (change[:5])

Test plan

• Unit tests pass (test_interpretation_common.py, test_reason_core.py, test_reason_misc.py)
• New functional test test_inconsistency_trace.py passes across regular, fp, and parallel modes
• Run inconsistency_example.py and verify new columns appear in CSV output
• Verify existing rule trace consumers (e.g., filter_and_sort_nodes) still work with the expanded tuple

🤖 Generated with Claude Code