Zotero Chunk RAG

# Phase 0: Dead Code Removal ## Overview Remove the column-count-based grouping code from `combination.py`, fix the vacuous accuracy default in `ground_truth.py`, and stub `combine_hypotheses()` with `NotImplementedError` until Phase 2 provides the per-divider voting replacement. **Entry state**: Pipeline functional with column-count grouping. **Exit state**: Pipeline broken for multi-hypothesis combination (single-hypothesis passthrough and empty-list handling still work). 8 tests that require multi-hypothesis combination will fail with `NotImplementedError` (expected — restored in Phase 2). --- ## Wave 0.1: Remove obsolete combination and metric code ### Task 0.1.1: Remove column-count grouping and fix vacuous accuracy default - **Description**: Remove `_select_best_column_group()` and `_group_confidence()` from `combination.py`. Stub the multi-hypothesis path in `combine_hypotheses()` with `NotImplementedError`. Change the vacuous `else 100.0` default in `ground_truth.py` to `else 0.0`. Delete `TestColumnGrouping` tests. 8 tests that call `combine_hypotheses()` with 2+ hypotheses will fail with `NotImplementedError` until Phase 2 restores them. - **Files to modify**: - `src/zotero_chunk_rag/feature_extraction/combination.py`: - Delete `_select_best_column_group()` (lines 157-213) - Delete `_group_confidence()` (lines 216-222) - In `combine_hypotheses()`: replace the multi-hypothesis body (everything after the single-hypothesis passthrough, starting at line 97) with `raise NotImplementedError("awaiting per-divider voting rewrite")` - Keep empty-list handling (lines 53-75) and single-hypothesis passthrough (lines 77-95) unchanged - Keep all remaining functions unchanged: `_make_passthrough_trace()`, `_compute_spatial_precision()`, `_expand_point_estimate()`, `_merge_overlapping()`, `_combine_axis()` - `src/zotero_chunk_rag/feature_extraction/ground_truth.py`: - Line 686: change `else 100.0` to `else 0.0` in the `cell_accuracy_pct` computation - `tests/test_combination.py`: - Delete `TestColumnGrouping` class entirely - Delete `_make_hyp()` helper function (only used by `TestColumnGrouping`) - Remove `_select_best_column_group` from the import block - The following 8 tests will fail with `NotImplementedError` after this change (expected — Phase 2 restores them): - `TestCombineHypotheses::test_two_methods_agree` - `TestCombineHypotheses::test_three_methods_one_disagrees` - `TestCombineHypotheses::test_row_and_col_independent` - `TestAcceptanceThreshold::test_low_confidence_rejected` - `TestAcceptanceThreshold::test_percentile_based_acceptance` - `TestAcceptanceThreshold::test_single_cluster_always_accepted` - `TestCombinationTrace::test_multi_hypothesis_trace_structure` - `TestCombinationTrace::test_axis_trace_input_points_match` - **Tests** (all should pass after changes): - `tests/test_combination.py::TestCombineHypotheses::test_empty_hypotheses_list` — assert returns empty consensus - `tests/test_combination.py::TestCombineHypotheses::test_single_hypothesis_passthrough` — assert returns passthrough - `tests/test_combination.py::TestAcceptanceThreshold::test_single_high_confidence_accepted` — single hypothesis, still works - `tests/test_combination.py::TestCombinationTrace::test_trace_false_returns_hypothesis_only` — single hypothesis - `tests/test_combination.py::TestCombinationTrace::test_trace_true_returns_tuple` — single hypothesis - `tests/test_combination.py::TestCombinationTrace::test_empty_hypotheses_trace` — empty input - `tests/test_combination.py::TestCombinationTrace::test_single_hypothesis_passthrough_trace` — single hypothesis - `tests/test_combination.py::TestExpandPointEstimates::*` — all unchanged - `tests/test_combination.py::TestOverlapMerge::*` — all unchanged - `tests/test_combination.py::TestSpatialPrecision::*` — all unchanged - 8 tests listed above fail with `NotImplementedError` (expected — restored in Phase 2) - **Acceptance criteria**: - `_select_best_column_group` and `_group_confidence` no longer exist in `combination.py` - `grep -r "_select_best_column_group\|_group_confidence" src/` returns nothing - `combine_hypotheses([h1, h2], ctx)` raises `NotImplementedError` - `combine_hypotheses([], ctx)` and `combine_hypotheses([h1], ctx)` still work correctly - `ground_truth.py` cell_accuracy_pct default is `else 0.0` - `TestColumnGrouping` class no longer exists - All other tests pass; 8 tests listed above fail with `NotImplementedError` (restored in Phase 2)