Zotero Chunk RAG

# Phase 5: Legacy Reference Review ## Overview Audit the entire repository for any remaining references to removed code, old algorithm descriptions, or stale documentation. No legacy references are acceptable in any form. **Depends on**: Phase 4 **Entry state**: Pipeline passes all targets. Code is correct but may have stale references. **Exit state**: Clean repository with no references to removed functions, old algorithms, or superseded spec files. --- ## Wave 5.1: Full legacy audit ### Task 5.1.1: Search and remove all stale references - **Description**: Search the entire repository for references to removed or changed code. Fix or remove every match. **Search targets** (grep repo-wide for each): - `_select_best_column_group` — removed function - `_group_confidence` — removed function - `"else 100.0"` — old vacuous default (now `else 0.0`) - `"column count grouping"` or `"column-count grouping"` — old algorithm description - `"column-count-based"` — old algorithm description - `"Q1 percentile"` or `"25th percentile"` in combination context — old acceptance logic - `normalize_method_confidence` — removed toggle (never implemented) - `reevaluate_accuracy` — removed script (never created) - `extract_with_all_boundaries` — removed method (already checked, but verify) **Files to check specifically**: - `CLAUDE.md` — update architecture notes if they reference old combination algorithm - `MEMORY.md` — update if references old algorithm - `spec/plan.md` — update dependency graph to reflect actual execution order - `spec/pipeline_operators_guide.md` — update if references old combination logic - All files in `src/zotero_chunk_rag/feature_extraction/` — check docstrings, comments - All files in `tests/` — check test comments, docstrings Also verify: - No dead imports in `combination.py` (all imports are used) - No dead imports in `ground_truth.py` - No references to `table_extraction` package (old name, already checked by existing test but verify no new references crept in) - **Files to modify**: Any file containing stale references (determined by search) - **Tests**: - `tests/test_feature_extraction/test_integration.py::TestDocs::test_claude_md_no_table_extraction_refs` — existing test, should still pass - `tests/test_feature_extraction/test_integration.py::TestDocs::test_claude_md_no_figure_extraction_ref` — existing test, should still pass - Manual verification: `grep -r "_select_best_column_group\|_group_confidence" src/ tests/` returns 0 matches - Manual verification: `grep -r "else 100.0" src/zotero_chunk_rag/feature_extraction/ground_truth.py` returns 0 matches - **Acceptance criteria**: - `grep -r "_select_best_column_group" .` returns 0 matches repo-wide - `grep -r "_group_confidence" .` returns 0 matches repo-wide - `grep -r "column.count.group" src/ tests/` returns 0 matches (case-insensitive) - `grep -r "normalize_method_confidence" .` returns 0 matches repo-wide - `CLAUDE.md` accurately describes the current combination algorithm (per-divider voting) - `MEMORY.md` accurately describes the current architecture - No dead imports in any modified source file - All existing tests pass