"""Formal stage scrap-rate evidence, coverage, report, and chart contracts.""" import sys import unittest from pathlib import Path sys.path.insert(0, str(Path(__file__).resolve().parents[1] / 'scripts')) from analysis_visuals import _comparison_charts, _trend_chart_from_points, build_visuals from nao_agent_report import _build_trend_report, build_noncomparison_report from nao_analysis_coverage import formal_metric_coverage, semantic_metric_gaps from nao_evidence_finalizer import _formal_metric_report from nao_evidence_report import build_report RELEASE = 'release-scrap-rate' QUESTION = '查询2026年9月黑片报废率' PERIOD = {'year': 2026, 'month': 9} COMPARISON_PERIOD = {'year': 2026, 'month': 8} def task(question=QUESTION, *, period=PERIOD, comparison_period=None, stages=None, fields=None, trend=False, visual='none'): return { 'version': 1, 'question': question, 'kind': 'metrics', 'filters': {'product': '产品甲'}, 'period': period, 'comparisonPeriod': comparison_period, 'stages': stages if stages is not None else ['black'], 'fields': fields if fields is not None else ['scrapRate'], 'dimensions': [], 'trendRequired': trend, 'comparisonRequired': comparison_period is not None, 'freshQuery': True, 'verifySource': False, 'fullDetails': False, 'needsClarification': False, 'visualPreference': visual, 'visualType': None, 'responseStyle': 'standard', 'periodCoverage': 'available', 'historyMode': 'none', } def metric_doc(*, value=0.125, month=9, filters=None, stage='black', rate=None, evidence='A', rows=2): period = {'year': 2026, 'month': month} summary_stage = {'scrapRate': value, 'totalWeight': 100, 'scrapWeight': value * 100 if isinstance(value, (int, float)) and value >= 0 else 0} if rate is not None: summary_stage['rate'] = rate return { 'tool': 'quality_data', 'mode': 'query', 'releaseId': RELEASE, 'evidenceRef': 'Q' + evidence * 8, 'result': { 'kind': 'quality_metrics', 'releaseId': RELEASE, 'period': period, 'filters': filters or {'product': '产品甲'}, 'view': 'overall', 'summary': {stage: summary_stage}, 'coverage': {'sourceMonths': [{'year': 2026, 'month': month, 'blackRows': rows if stage == 'black' else 0, 'brightRows': rows if stage == 'bright' else 0}]}, }, } def comparison_doc(*, black_values=(0.05, 0.0), release=RELEASE, filters=None, black_rows=(2, 2), evidence='B'): def snapshot(month, value, rows): return { 'period': {'year': 2026, 'month': month}, 'compositeRate': 0.94, 'black': {'scrapRate': value, 'rate': 0.95, 'totalWeight': 100, 'scrapWeight': 5}, 'bright': {'scrapRate': 0.06, 'rate': 0.96, 'totalWeight': 90, 'scrapWeight': 4}, 'coverage': {'sourceMonths': [{'year': 2026, 'month': month, 'blackRows': rows, 'brightRows': 2}]}, } result = { 'kind': 'quality_comparison', 'releaseId': release, 'filterBinding': filters or {'product': '产品甲'}, 'comparison': snapshot(8, black_values[0], black_rows[0]), 'current': snapshot(9, black_values[1], black_rows[1]), 'delta': {'available': True, 'percentagePoints': 0}, 'stageRateChangesPP': {'black': 0, 'bright': 0}, } return {'tool': 'quality_data', 'mode': 'compare', 'releaseId': release, 'evidenceRef': 'Q' + evidence * 8, 'result': result} class ScrapRateEvidenceTest(unittest.TestCase): def context(self, selected_task): return {'releaseId': RELEASE, 'period': selected_task['period'], 'comparisonPeriod': selected_task['comparisonPeriod'], 'filters': selected_task['filters'], '_agentTask': selected_task} def test_single_period_uses_scrap_rate_directly_and_keeps_zero_and_over_one(self): for raw, expected in ((0, '0.00%'), (1.25, '125.00%')): selected = task() report = build_noncomparison_report( [metric_doc(value=raw, rate=None)], self.context(selected), QUESTION) self.assertEqual([fact['label'] for fact in report['facts']], ['黑片报废率']) self.assertIn(expected, report['facts'][0]['text']) self.assertTrue(report['coverageSufficient']) def test_requested_numerator_and_denominator_are_displayed_as_independent_formal_values(self): question='黑片报废率是多少,同时列出分子分母' selected=task(question,fields=['scrapRate','scrapWeight','totalWeight']) doc=metric_doc(value=.175,rate=.99) doc['result']['summary']['black'].update(totalWeight=200,scrapWeight=35) report=build_noncomparison_report([doc],self.context(selected),question) rendered='\n'.join(report['sections']) for expected in ('17.50%','35.00','200.00'): self.assertIn(expected,rendered) self.assertEqual([f['label'] for f in report['facts']],['黑片报废率','黑片报废重量','黑片投入重量']) self.assertTrue(report['coverageSufficient']) self.assertNotIn('1.00%',rendered) def test_missing_requested_denominator_is_incomplete_even_when_formal_rate_exists(self): question='黑片报废率及报废总量和修正总重' selected=task(question,fields=['scrapRate','scrapWeight','totalWeight']) doc=metric_doc();doc['result']['summary']['black'].pop('totalWeight') report=build_noncomparison_report([doc],self.context(selected),question) self.assertFalse(report['coverageSufficient']) self.assertTrue(any('重量' in gap for gap in report['missingEvidence'])) def test_missing_or_negative_scrap_rate_is_not_treated_as_zero_or_clamped(self): for raw in (None, -0.01): selected = task() report = build_noncomparison_report( [metric_doc(value=raw, rate=0.99)], self.context(selected), QUESTION) self.assertIn('未提供', report['facts'][0]['text']) self.assertFalse(report['coverageSufficient']) self.assertEqual(semantic_metric_gaps( [metric_doc(value=None)], QUESTION, self.context(task())), ['未取得指定期间、范围的黑片报废率正式指标或来源覆盖。']) def test_scrap_rate_coverage_does_not_require_qualified_rate_or_other_stage(self): selected = task() doc = metric_doc(value=0.02, rate=None) self.assertEqual(semantic_metric_gaps([doc], QUESTION, self.context(selected)), []) report = build_noncomparison_report([doc], self.context(selected), QUESTION) self.assertEqual([fact['label'] for fact in report['facts']], ['黑片报废率']) self.assertNotIn('亮片', '\n'.join(fact['text'] for fact in report['facts'])) self.assertNotIn('合格率', '\n'.join(fact['text'] for fact in report['facts'])) def test_scope_and_period_mismatch_are_rejected(self): selected = task() wrong_scope = metric_doc(value=0.02, filters={'product': '产品乙'}) wrong_period = metric_doc(value=0.02, month=8) self.assertTrue(semantic_metric_gaps([wrong_scope], QUESTION, self.context(selected))) self.assertTrue(semantic_metric_gaps([wrong_period], QUESTION, self.context(selected))) def test_two_period_formal_contract_and_body_use_only_scrap_rate(self): question = '只看产品甲,2026年9月与8月黑片报废率' selected = task(question, period=PERIOD, comparison_period=COMPARISON_PERIOD) context = self.context(selected) doc = comparison_doc(black_values=(1.25, 0.0)) contract = formal_metric_coverage([doc], question, context) self.assertEqual(contract['fields'], ['scrapRate']) self.assertEqual(contract['missingEvidence'], []) self.assertEqual([item['value'] for item in contract['metrics']], [0.0, 1.25]) report = _formal_metric_report(contract) text = '\n'.join(report['sections']) + '\n' + '\n'.join(item['text'] for item in report['facts']) self.assertIn('125.00%', text) self.assertIn('0.00%', text) self.assertNotIn('合格率', text) wrong_scope = comparison_doc(filters={'product': '产品乙'}) rejected = formal_metric_coverage([wrong_scope], question, context) self.assertTrue(rejected['missingEvidence']) comparison_question = '对比2026年9月与8月黑片报废率' selected = task(comparison_question, period=PERIOD, comparison_period=COMPARISON_PERIOD, stages=['black']) body = build_report([doc], self.context(selected), comparison_question) body_text = '\n'.join(body['sections']) + '\n' + '\n'.join(fact['text'] for fact in body['facts']) self.assertTrue(body['coverageSufficient'], body['missingEvidence']) self.assertIn('125.00%', body_text) self.assertIn('0.00%', body_text) self.assertIn('黑片报废率', body_text) self.assertNotIn('亮片报废率', body_text) def test_formal_scrap_weight_contract_rejects_negative_backend_weight(self): question = '只看产品甲,2026年9月与8月黑片报废重量' selected = task(question, period=PERIOD, comparison_period=COMPARISON_PERIOD, fields=['scrapWeight']) document = comparison_doc() document['result']['current']['black']['scrapWeight'] = -0.01 contract = formal_metric_coverage([document], question, self.context(selected)) self.assertTrue(contract['missingEvidence']) invalid = [metric for metric in contract['metrics'] if metric['field'] == 'scrapWeight' and metric['value'] < 0] self.assertEqual(len(invalid), 1) self.assertFalse(invalid[0]['covered']) def test_omitted_stage_scrap_rate_stays_two_stage_not_a_composite(self): question = '查询2026年9月产品甲报废率' selected = task(question, stages=[]) context = self.context(selected) doc = metric_doc(value=0.04) doc['result']['summary']['bright'] = {'scrapRate': 0.03, 'totalWeight': 90, 'scrapWeight': 2.7} doc['result']['coverage']['sourceMonths'][0]['brightRows'] = 2 self.assertEqual(semantic_metric_gaps([doc], question, context), []) report = build_noncomparison_report([doc], context, question) self.assertEqual({fact['label'] for fact in report['facts']}, {'黑片报废率', '亮片报废率'}) self.assertFalse(any('综合报废率' in fact['text'] for fact in report['facts'])) def test_monthly_trend_uses_server_values_and_preserves_same_value(self): question = '查询2026年黑片报废率趋势' selected = task(question, period={'year': 2026}, trend=True, visual='required') trend = [] source_months = [] for month in (1, 2, 3): trend.append({'period': f'2026-{month:02d}', 'black': {'scrapRate': 0.08, 'totalWeight': 100, 'scrapWeight': 8}, 'coverage': {'blackRows': 2, 'brightRows': 0}}) source_months.append({'year': 2026, 'month': month, 'blackRows': 2, 'brightRows': 0}) doc = metric_doc(value=0.08) doc['result'].update(period={'year': 2026}, trend=trend, trendCount=3, trendTruncated=False, coverage={'sourceMonths': source_months}) report = _build_trend_report([doc], self.context(selected), question) self.assertTrue(report['coverageSufficient'], report['missingEvidence']) self.assertEqual([fact['label'] for fact in report['facts']], ['黑片报废率月度趋势']) self.assertIn('1月8.00%', report['facts'][0]['text']) self.assertIn('3月8.00%', report['facts'][0]['text']) self.assertIn('持平', report['facts'][0]['text']) self.assertNotIn('合格率', report['facts'][0]['text']) def test_trend_visual_uses_scrap_rate_contract_and_has_no_good_rate_series(self): points = [ {'period': f'2026-0{month}', 'snapshot': { 'black': {'scrapRate': value, 'rate': 0.95, 'totalWeight': 100, 'scrapWeight': value * 100}, 'coverage': {'blackRows': 2, 'brightRows': 0}}} for month, value in ((1, 0.0), (2, 1.25), (3, 0.0)) ] charts = _trend_chart_from_points( points, release_id=RELEASE, scope={'product': '产品甲'}, turn_id='turn-scrap', evidence_refs=['QAAAAAAAA'], question='黑片报废率趋势', supplied_trend=True, coverage_truncated=False, semantic_fields={'scrapRate'}, semantic_stages=['black']) self.assertEqual([chart['metric'] for chart in charts], ['black_scrap_rate']) self.assertEqual(charts[0]['unit'], 'ratio') self.assertEqual(charts[0]['requiredFields'], ['black_scrap_weight', 'black_total_weight']) self.assertEqual(charts[0]['series'][0]['values'], [0.0, 1.25, 0.0]) def test_comparison_chart_is_stage_specific_and_uses_scrap_authorization(self): document = comparison_doc(black_values=(1.25, 0.0)) charts = _comparison_charts( document, document['result'], 'turn-compare', '', requested_stage='black', semantic_fields={'scrapRate'}) self.assertEqual([chart['metric'] for chart in charts], ['black_scrap_rate']) self.assertEqual(charts[0]['unit'], 'ratio') self.assertEqual(charts[0]['requiredFields'], ['black_scrap_weight', 'black_total_weight']) self.assertEqual(charts[0]['series'][0]['values'], [1.25, 0.0]) self.assertEqual(charts[0]['type'], 'bar') def test_comparison_chart_rejects_missing_coverage_and_semantic_output_has_no_extra_rates(self): document = comparison_doc(black_rows=(0, 2)) charts = _comparison_charts(document, document['result'], 'turn-compare', '', requested_stage='black', semantic_fields={'scrapRate'}) self.assertEqual(charts, []) question = '对比2026年9月与8月黑片报废率' selected = task(question, period=PERIOD, comparison_period=COMPARISON_PERIOD, visual='required') context = self.context(selected) visuals = build_visuals([comparison_doc()], question, 'turn-semantic', context=context)['visuals'] self.assertEqual([chart['metric'] for chart in visuals], ['black_scrap_rate']) both_question = '对比2026年9月与8月黑片和亮片报废率' both_task = task(both_question, period=PERIOD, comparison_period=COMPARISON_PERIOD, stages=['black', 'bright'], visual='required') both_visuals = build_visuals([comparison_doc()], both_question, 'turn-both', context=self.context(both_task))['visuals'] self.assertEqual({chart['metric'] for chart in both_visuals}, {'black_scrap_rate', 'bright_scrap_rate'}) plain_question = '比较黑片报废率' plain_visuals = build_visuals([comparison_doc()], plain_question, 'turn-plain')['visuals'] self.assertEqual([chart['metric'] for chart in plain_visuals], ['black_scrap_rate']) if __name__ == '__main__': unittest.main()