优化:未分类连续维护与柏拉图预汇总,同步已发布修复
This commit is contained in:
@@ -78,6 +78,110 @@ class EchoColumnApi(FakeApi):
|
||||
|
||||
|
||||
class SourceColumnProtocolTests(unittest.TestCase):
|
||||
def test_single_product_combined_metric_uses_one_filtered_combined_ranking_node(self):
|
||||
product = 'PX-73X4.2XR26.4-VariantB'
|
||||
question = f'2026年9月{product}的综合合格率和黑片亮片合计报废总重是多少?只用文字回答。'
|
||||
plan = {
|
||||
'goal': '查询指定完整规格的综合合格率和黑片、亮片合计报废总重',
|
||||
'kind': 'metrics', 'filterUpdates': {'product': product},
|
||||
'period': {'year': 2026, 'month': 9}, 'comparisonPeriod': None,
|
||||
'comparisonRequired': False, 'trendRequired': False,
|
||||
'dimensions': ['product'], 'dimensionStages': {'product': []},
|
||||
'stages': [], 'fields': ['rate', 'scrapWeight'],
|
||||
'freshQuery': True, 'verifySource': False, 'fullDetails': False,
|
||||
'needsClarification': False, 'clarification': '',
|
||||
'visualPreference': 'none', 'visualType': None,
|
||||
'queryPlan': [{
|
||||
'nodeId': 'single_product', 'purpose': 'ranking', 'dimension': 'product',
|
||||
'limit': 1, 'fields': ['rate', 'scrapWeight'], 'stages': [],
|
||||
'selector': {'kind': 'base'}, 'visualType': None,
|
||||
'visualPreference': 'none', 'comparisonRequired': False,
|
||||
}],
|
||||
}
|
||||
review = {'aligned': True, 'requiresEvidence': True,
|
||||
'requiresSourceVerification': False, 'requiresAllSourceRows': False,
|
||||
'requiresTrend': False, 'requiresComparison': False}
|
||||
for protocol, replies in (
|
||||
({'version': 1}, [plan, review]),
|
||||
({'version': 1, 'autonomousEnabled': True}, [plan])):
|
||||
with self.subTest(autonomous=protocol.get('autonomousEnabled', False)):
|
||||
context = {'_protocol': protocol}
|
||||
with patch('nao_agent_task._model_json', side_effect=replies) as model:
|
||||
task = plan_task(base_url='https://offline.invalid/v1', api_key='unused', model='m',
|
||||
question=question, context=context, cancel_event=threading.Event())
|
||||
self.assertEqual(task['kind'], 'metrics')
|
||||
self.assertEqual(task['filters']['product'], product)
|
||||
self.assertEqual(task['dimensions'], ['product'])
|
||||
self.assertEqual(task['fields'], ['rate', 'scrapWeight'])
|
||||
self.assertEqual(task['stages'], [])
|
||||
self.assertEqual(task['dimensionStages']['product'], [])
|
||||
self.assertEqual(len(task['queryPlan']), 1)
|
||||
node = task['queryPlan'][0]
|
||||
self.assertEqual((node['purpose'], node['dimension'], node['limit']),
|
||||
('ranking', 'product', 1))
|
||||
self.assertEqual(node['fields'], ['rate', 'scrapWeight'])
|
||||
self.assertEqual(node['stages'], [])
|
||||
self.assertEqual(node['visualPreference'], 'none')
|
||||
self.assertIn('limit=1', model.call_args_list[0].kwargs['system'])
|
||||
self.assertIn('合计算子', model.call_args_list[0].kwargs['system'])
|
||||
|
||||
from nao_protocol_execution import _scoped_arguments
|
||||
args = _scoped_arguments({'releaseId': 'release-test'}, {'task': task}, node)
|
||||
self.assertEqual(len(args), 1)
|
||||
self.assertIsNone(args[0]['stage'])
|
||||
self.assertEqual(args[0]['metric'], 'scrapWeight')
|
||||
|
||||
def test_unqualified_bad_top_stays_combined_product_ranking_under_exact_product_filter(self):
|
||||
product = 'DYN-SPEC-0.795+0.937X1.498X7'
|
||||
question = f'2026年9月只看规格{product}的不良TOP5。'
|
||||
plan = {
|
||||
'goal': '按用户给出的完整产品规格范围列出不良TOP5',
|
||||
'kind': 'analysis', 'filterUpdates': {'product': product},
|
||||
'period': {'year': 2026, 'month': 9}, 'comparisonPeriod': None,
|
||||
'comparisonRequired': False, 'trendRequired': False,
|
||||
'dimensions': ['product'], 'dimensionStages': {'product': []},
|
||||
'stages': [], 'fields': ['rate', 'scrapWeight'],
|
||||
'freshQuery': True, 'verifySource': False, 'fullDetails': False,
|
||||
'needsClarification': False, 'clarification': '',
|
||||
'visualPreference': 'auto', 'visualType': 'bar',
|
||||
'queryPlan': [{
|
||||
'nodeId': 'product_top', 'purpose': 'ranking', 'dimension': 'product',
|
||||
'limit': 5, 'fields': ['rate', 'scrapWeight'], 'stages': [],
|
||||
'selector': {'kind': 'base'}, 'visualType': 'bar',
|
||||
'visualPreference': 'auto', 'comparisonRequired': False,
|
||||
}],
|
||||
}
|
||||
review = {'aligned': True, 'requiresEvidence': True,
|
||||
'requiresSourceVerification': False, 'requiresAllSourceRows': False,
|
||||
'requiresTrend': False, 'requiresComparison': False}
|
||||
for protocol, replies in (
|
||||
({'version': 1}, [plan, review]),
|
||||
({'version': 1, 'autonomousEnabled': True}, [plan])):
|
||||
with self.subTest(autonomous=protocol.get('autonomousEnabled', False)):
|
||||
with patch('nao_agent_task._model_json', side_effect=replies) as model:
|
||||
task = plan_task(base_url='https://offline.invalid/v1', api_key='unused', model='m',
|
||||
question=question, context={'_protocol': protocol}, cancel_event=threading.Event())
|
||||
self.assertEqual(task['kind'], 'analysis')
|
||||
self.assertEqual(task['filters']['product'], product)
|
||||
self.assertEqual(task['dimensions'], ['product'])
|
||||
self.assertEqual(task['fields'], ['rate', 'scrapWeight'])
|
||||
self.assertEqual(task['stages'], [])
|
||||
self.assertEqual(task['dimensionStages']['product'], [])
|
||||
self.assertNotIn('缺陷', task['goal'])
|
||||
self.assertEqual(len(task['queryPlan']), 1)
|
||||
node = task['queryPlan'][0]
|
||||
self.assertEqual((node['purpose'], node['dimension'], node['limit']),
|
||||
('ranking', 'product', 5))
|
||||
self.assertEqual(node['stages'], [])
|
||||
self.assertNotIn('defects', [item['purpose'] for item in task['queryPlan']])
|
||||
self.assertIn('空榜', model.call_args_list[0].kwargs['system'])
|
||||
|
||||
from nao_protocol_execution import _scoped_arguments
|
||||
args = _scoped_arguments({'releaseId': 'release-test'}, {'task': task}, node)
|
||||
self.assertEqual(len(args), 1)
|
||||
self.assertIsNone(args[0]['stage'])
|
||||
self.assertEqual(args[0]['metric'], 'scrapWeight')
|
||||
|
||||
def test_final_answer_selector_preserves_the_confirmed_column_scope(self):
|
||||
from nao_evidence_finalizer import _agent_task_summary
|
||||
task=protocol_task()
|
||||
@@ -97,7 +201,7 @@ class SourceColumnProtocolTests(unittest.TestCase):
|
||||
catalog = {'sourceFields': [canonical_product, *SOURCE_FIELDS[1:]], 'departments': [], 'groups': []}
|
||||
task = normalize_task(task_value(kind='members', dimensions=[], fields=[], stages=['black', 'bright'],
|
||||
period={'year': 2026, 'month': 3}, comparisonRequired=False, trendRequired=False,
|
||||
memberFields=['product'], columnFilters=[]), '列出客户规格清单',
|
||||
memberFields=['product'], columnFilters=[]), '列出产品清单',
|
||||
{'_protocol': {'version': 1}, '_businessCatalog': catalog})
|
||||
self.assertEqual(task['memberFields'], ['product'])
|
||||
self.assertEqual(task['sourceFieldLabels'], {'product': '产品'})
|
||||
@@ -110,6 +214,134 @@ class SourceColumnProtocolTests(unittest.TestCase):
|
||||
columnFilters=[{'field': 'product', 'stages': ['black'], 'operator': 'in', 'values': ['18X15X4']}]),
|
||||
'按产品筛选指标', {'_protocol': {'version': 1}, '_businessCatalog': catalog})
|
||||
|
||||
def test_member_fields_must_match_the_requested_entity_not_lookalike_or_filter_columns(self):
|
||||
canonical_product = {**SOURCE_FIELDS[0], 'type': 'dimension'}
|
||||
product_spec = source_field('src_dynamic_product_spec_0', '产品规格', '黑片规格', '亮片规格')
|
||||
drawing = source_field('src_dynamic_drawing_1', '产品图号', '图号-黑片', '图号-亮片')
|
||||
customer_code = source_field('src_dynamic_customer_2', '客户代码', '客户编码-黑片', '客户编码-亮片')
|
||||
customer_spec = source_field('src_dynamic_customer_spec_3', '客户规格', '客户规格-黑片', '客户规格-亮片')
|
||||
catalog = {'sourceFields': [canonical_product, product_spec, drawing, customer_code, customer_spec],
|
||||
'extraFields': [{'key': 'customerCode', 'label': '客户代码'}],
|
||||
'departments': [], 'groups': []}
|
||||
context = {'_protocol': {'version': 1}, '_businessCatalog': catalog}
|
||||
question = '800001客户都有哪些规格?'
|
||||
wrong = task_value(kind='members', dimensions=[], fields=[], stages=['black', 'bright'],
|
||||
period={'year': 2026, 'month': 9}, comparisonRequired=False, trendRequired=False,
|
||||
memberFields=[drawing['key'], customer_code['key']], columnFilters=[],
|
||||
filterUpdates={'extraFilters': {'customerCode': ['800001']}})
|
||||
with self.assertRaises(TaskUnavailable) as raised:
|
||||
normalize_task(wrong, question, context)
|
||||
self.assertEqual(raised.exception.field, 'memberFields')
|
||||
self.assertIn('产品规格', str(raised.exception))
|
||||
|
||||
correct = normalize_task({**wrong, 'memberFields': [product_spec['key']]}, question, context)
|
||||
self.assertEqual(correct['memberFields'], [product_spec['key']])
|
||||
self.assertEqual(correct['sourceFieldLabels'], {product_spec['key']: '产品规格'})
|
||||
self.assertEqual(correct['filters']['extraFilters'], {'customerCode': ['800001']})
|
||||
|
||||
drawing_task = {**wrong, 'memberFields': [drawing['key']]}
|
||||
drawing_result = normalize_task(drawing_task, '800001客户的产品图号有哪些?', context)
|
||||
self.assertEqual(drawing_result['memberFields'], [drawing['key']])
|
||||
|
||||
without_product_spec = {'sourceFields': [canonical_product, drawing, customer_code],
|
||||
'extraFields': [{'key': 'customerCode', 'label': '客户代码'}],
|
||||
'departments': [], 'groups': []}
|
||||
missing = normalize_task(wrong, question,
|
||||
{'_protocol': {'version': 1}, '_businessCatalog': without_product_spec})
|
||||
self.assertTrue(missing['needsClarification'])
|
||||
self.assertEqual(missing['memberFields'], [])
|
||||
self.assertIn('没有可用于', missing['clarification'])
|
||||
|
||||
customer_spec_task = task_value(kind='members', dimensions=[], fields=[], stages=['black', 'bright'],
|
||||
period={'year': 2026, 'month': 9}, comparisonRequired=False, trendRequired=False,
|
||||
memberFields=[customer_spec['key']], columnFilters=[])
|
||||
customer_spec_result = normalize_task(customer_spec_task, '800001客户有哪些客户规格?', context)
|
||||
self.assertEqual(customer_spec_result['memberFields'], [customer_spec['key']])
|
||||
with self.assertRaises(TaskUnavailable) as rejected:
|
||||
normalize_task({**customer_spec_task, 'memberFields': [product_spec['key']]}, '800001客户有哪些客户规格?', context)
|
||||
self.assertEqual(rejected.exception.field, 'memberFields')
|
||||
|
||||
def test_ambiguous_member_spec_columns_force_clarification_without_a_query(self):
|
||||
primary = source_field('src_dynamic_product_spec_1', '产品规格', '规格-黑片', '规格-亮片')
|
||||
secondary = source_field('src_dynamic_finished_spec_2', '成品规格', '成品规格-黑片', '成品规格-亮片')
|
||||
material_spec = source_field('src_dynamic_material_spec_3', '原材料规格', '原料规格-黑片', '原料规格-亮片')
|
||||
version = source_field('src_dynamic_spec_version_4', '规格版本', '规格版本-黑片', '规格版本-亮片')
|
||||
catalog = {'sourceFields': [primary, secondary, material_spec, version], 'departments': [], 'groups': []}
|
||||
planned = normalize_task(task_value(kind='members', dimensions=[], fields=[], stages=['black', 'bright'],
|
||||
period={'year': 2026, 'month': 9}, memberFields=[primary['key']], columnFilters=[]), '列出本期规格清单',
|
||||
{'_protocol': {'version': 1}, '_businessCatalog': catalog})
|
||||
self.assertTrue(planned['needsClarification'])
|
||||
self.assertEqual(planned['memberFields'], [])
|
||||
self.assertIn('多个', planned['clarification'])
|
||||
|
||||
explicit = normalize_task(task_value(kind='members', dimensions=[], fields=[], stages=['black', 'bright'],
|
||||
period={'year': 2026, 'month': 9}, memberFields=[secondary['key']], columnFilters=[]),
|
||||
'列出本期成品规格清单', {'_protocol': {'version': 1}, '_businessCatalog': catalog})
|
||||
self.assertFalse(explicit['needsClarification'])
|
||||
self.assertEqual(explicit['memberFields'], [secondary['key']])
|
||||
|
||||
def test_member_semantics_keep_multiple_requested_entities_and_ignore_lookalike_labels(self):
|
||||
product_spec = source_field('src_dynamic_product_spec_5', '产品规格', '产品规格-黑片', '产品规格-亮片')
|
||||
customer_spec = source_field('src_dynamic_customer_spec_6', '客户规格', '客户规格-黑片', '客户规格-亮片')
|
||||
material_spec = source_field('src_dynamic_material_spec_7', '原材料规格', '原料规格-黑片', '原料规格-亮片')
|
||||
spec_version = source_field('src_dynamic_version_8', '规格版本', '规格版本-黑片', '规格版本-亮片')
|
||||
catalog = {'sourceFields': [product_spec, customer_spec, material_spec, spec_version],
|
||||
'departments': [], 'groups': []}
|
||||
value = task_value(kind='members', dimensions=[], fields=[], stages=['black', 'bright'],
|
||||
period={'year': 2026, 'month': 9}, memberFields=[customer_spec['key'], product_spec['key']], columnFilters=[])
|
||||
planned = normalize_task(value, '列出成品规格,并显示客户规格',
|
||||
{'_protocol': {'version': 1}, '_businessCatalog': catalog})
|
||||
self.assertFalse(planned['needsClarification'])
|
||||
self.assertEqual(set(planned['memberFields']), {product_spec['key'], customer_spec['key']})
|
||||
|
||||
negated = normalize_task({**value, 'memberFields': [product_spec['key']]},
|
||||
'列出成品规格,不要显示客户规格',
|
||||
{'_protocol': {'version': 1}, '_businessCatalog': catalog})
|
||||
self.assertFalse(negated['needsClarification'])
|
||||
self.assertEqual(negated['memberFields'], [product_spec['key']])
|
||||
|
||||
raw_material = source_field('src_dynamic_raw_material_spec_9', '原材料规格', '原材料规格黑片', '原材料规格亮片')
|
||||
explicit_label_catalog = {'sourceFields': [product_spec, raw_material, spec_version],
|
||||
'departments': [], 'groups': []}
|
||||
raw_material_result = normalize_task({**value, 'memberFields': [raw_material['key']]},
|
||||
'列出原材料规格',
|
||||
{'_protocol': {'version': 1}, '_businessCatalog': explicit_label_catalog})
|
||||
self.assertFalse(raw_material_result['needsClarification'])
|
||||
self.assertEqual(raw_material_result['memberFields'], [raw_material['key']])
|
||||
|
||||
only_product_spec_catalog = {'sourceFields': [product_spec, material_spec, spec_version],
|
||||
'departments': [], 'groups': []}
|
||||
single = normalize_task({**value, 'memberFields': [product_spec['key']]}, '列出本期规格清单',
|
||||
{'_protocol': {'version': 1}, '_businessCatalog': only_product_spec_catalog})
|
||||
self.assertEqual(single['memberFields'], [product_spec['key']])
|
||||
|
||||
def test_member_semantic_repair_uses_this_turn_catalog_and_does_not_add_another_reviewer(self):
|
||||
canonical_product = {**SOURCE_FIELDS[0], 'type': 'dimension'}
|
||||
product_spec = source_field('src_run_product_spec_7', '产品规格', '产品规格黑片', '产品规格亮片')
|
||||
drawing = source_field('src_run_drawing_9', '产品图号', '图号-黑片', '图号-亮片')
|
||||
customer_code = source_field('src_run_code_8', '客户代码', '客户编码-黑片', '客户编码-亮片')
|
||||
catalog = {'sourceFields': [canonical_product, product_spec, drawing, customer_code],
|
||||
'extraFields': [{'key': 'customerCode', 'label': '客户代码'}],
|
||||
'departments': [], 'groups': []}
|
||||
question = '800001客户都有哪些规格?'
|
||||
context = {'_protocol': {'version': 1, 'autonomousEnabled': True}, '_businessCatalog': catalog}
|
||||
base = task_value(kind='members', goal='列出客户800001的产品规格清单', dimensions=[], fields=[],
|
||||
stages=['black', 'bright'], period={'year': 2026, 'month': 9}, comparisonRequired=False,
|
||||
trendRequired=False, memberFields=[drawing['key'], customer_code['key']], columnFilters=[],
|
||||
filterUpdates={'extraFilters': {'customerCode': ['800001']}}, visualPreference='none')
|
||||
repaired = {**base, 'memberFields': [product_spec['key']]}
|
||||
with patch('nao_agent_task._model_json', side_effect=[base, repaired]) as model:
|
||||
planned = plan_task(base_url='https://offline.invalid/v1', api_key='never-used', model='m',
|
||||
question=question, context=context, cancel_event=threading.Event())
|
||||
self.assertEqual(planned['memberFields'], [product_spec['key']])
|
||||
self.assertEqual(model.call_count, 2)
|
||||
repair_payload = model.call_args_list[1].kwargs['payload']
|
||||
self.assertEqual(repair_payload['validationDetails']['field'], 'memberFields')
|
||||
self.assertIn('产品图号', str(repair_payload['businessCatalog']['sourceFields']))
|
||||
self.assertIn('memberFields', model.call_args_list[0].kwargs['system'])
|
||||
self.assertIn('图号', model.call_args_list[0].kwargs['system'])
|
||||
self.assertIn('产品规格', model.call_args_list[0].kwargs['system'])
|
||||
|
||||
def test_column_filter_value_length_matches_backend_256_character_limit(self):
|
||||
rule = {'field': 'batchNumber', 'stages': ['black'], 'operator': 'in', 'values': ['X' * 256]}
|
||||
task = protocol_task(columnFilters=[rule])
|
||||
@@ -337,8 +569,11 @@ class SourceColumnProtocolTests(unittest.TestCase):
|
||||
catalog = {**CATALOG, 'extraFields': [{'key': 'customerCode'}]}
|
||||
context = bind_task({'_protocol': {'version': 1}, '_businessCatalog': catalog,
|
||||
'releaseId': RELEASE}, previous)
|
||||
product_spec = source_field('src_dynamic_new_task_spec', '产品规格', '产品规格黑片', '产品规格亮片')
|
||||
catalog['sourceFields'] = [*SOURCE_FIELDS, product_spec]
|
||||
context['_businessCatalog'] = catalog
|
||||
good = task_value(kind='members', dimensions=[], stages=['black', 'bright'], fields=[],
|
||||
memberFields=['product'], newTask=True, columnFilters=[],
|
||||
memberFields=[product_spec['key']], newTask=True, columnFilters=[],
|
||||
period={'year':2026,'month':9}, filterUpdates={'extraFilters':{'customerCode':['800001']}})
|
||||
bad = {**good, 'filterUpdates': {'customerCode':['800001']}}
|
||||
review = {'aligned':True, 'requiresEvidence':True, 'requiresSourceVerification':False,
|
||||
@@ -352,7 +587,7 @@ class SourceColumnProtocolTests(unittest.TestCase):
|
||||
self.assertTrue(repair['invalidPlan']['newTask'])
|
||||
self.assertEqual(planned['filters']['extraFilters'], {'customerCode':['800001']})
|
||||
self.assertEqual(planned['filters']['columnFilters'], [])
|
||||
self.assertEqual(planned['memberFields'], ['product'])
|
||||
self.assertEqual(planned['memberFields'], [product_spec['key']])
|
||||
self.assertEqual(planned['stages'], ['black','bright'])
|
||||
self.assertTrue(planned['newTask'])
|
||||
|
||||
|
||||
Reference in New Issue
Block a user