266 lines
17 KiB
JSON
266 lines
17 KiB
JSON
{
|
||
"schema_version": "domain_truth_harness_spec_v1",
|
||
"scenario_id": "agent_cashflow_profit_directness_20260523",
|
||
"domain": "autonomy_business_answer_contract",
|
||
"title": "AGENT | Cashflow vs profit directness pack",
|
||
"description": "Targeted AGENT replay for cashflow-vs-profit directness: colloquial earnings questions must produce a compact money answer, while broad overview remains available only when explicitly requested.",
|
||
"bindings": {
|
||
|
||
},
|
||
"steps": [
|
||
{
|
||
"step_id": "step_01_direct_money_earned_explicit_shape",
|
||
"title": "Direct money earned question asks for compact cashflow shape",
|
||
"question": "Сколько денег Альтернатива заработала за 2020 год? Ответь коротко: получили, заплатили, денежное нетто, это прибыль или нет.",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"47[\\s.]*628[\\s.]*853",
|
||
"43[\\s.]*763[\\s.]*351",
|
||
"3[\\s.]*865[\\s.]*501"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive",
|
||
"7\\s*136\\s*815|7136815",
|
||
"Комитет государственных услуг"
|
||
],
|
||
"criticality": "critical",
|
||
"semantic_tags": [
|
||
"cashflow_vs_profit",
|
||
"direct_answer",
|
||
"colloquial_earnings"
|
||
]
|
||
},
|
||
{
|
||
"step_id": "step_02_colloquial_money_earned",
|
||
"title": "Colloquial earned wording stays cashflow, not broad overview",
|
||
"question": "скока денег альтернатива заработала за 20 год?",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"47[\\s.]*628[\\s.]*853",
|
||
"43[\\s.]*763[\\s.]*351",
|
||
"3[\\s.]*865[\\s.]*501"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive",
|
||
"7\\s*136\\s*815|7136815",
|
||
"Комитет государственных услуг",
|
||
"Что проверить дальше"
|
||
],
|
||
"criticality": "critical",
|
||
"semantic_tags": [
|
||
"cashflow_vs_profit",
|
||
"slang_wording",
|
||
"no_overview_substitution"
|
||
]
|
||
},
|
||
{
|
||
"step_id": "step_03_profit_followup_after_cashflow",
|
||
"title": "Clean profit follow-up keeps distinction",
|
||
"question": "а это чистая прибыль?",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"7\\s*136\\s*815|7136815",
|
||
"90|91|99"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive"
|
||
],
|
||
"criticality": "critical",
|
||
"semantic_tags": [
|
||
"profit_vs_cashflow",
|
||
"followup_context"
|
||
]
|
||
},
|
||
{
|
||
"step_id": "step_04_money_plus_or_minus_followup",
|
||
"title": "Money plus/minus follow-up returns cashflow net",
|
||
"question": "а по деньгам тогда плюс или минус?",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"\\+",
|
||
"47[\\s.]*628[\\s.]*853",
|
||
"43[\\s.]*763[\\s.]*351",
|
||
"3[\\s.]*865[\\s.]*501"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive",
|
||
"7\\s*136\\s*815|7136815"
|
||
],
|
||
"criticality": "critical",
|
||
"semantic_tags": [
|
||
"cashflow_polarity",
|
||
"followup_context",
|
||
"no_profit_substitution"
|
||
]
|
||
},
|
||
{
|
||
"step_id": "step_05_explicit_business_overview_allowed",
|
||
"title": "Explicit adult overview may include business structure and next steps",
|
||
"question": "Теперь дай взрослый обзор за 2020 по компании: входящие, исходящие, нетто, топы, но банк в топах отдельно объясни как финансовый поток.",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"47[\\s.]*628[\\s.]*853",
|
||
"43[\\s.]*763[\\s.]*351",
|
||
"3[\\s.]*865[\\s.]*501",
|
||
"12[\\s.]*792[\\s.]*194",
|
||
"12[\\s.]*093[\\s.]*465",
|
||
"9[\\s.]*612[\\s.]*904"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive"
|
||
],
|
||
"criticality": "high",
|
||
"semantic_tags": [
|
||
"business_overview",
|
||
"bank_flow_boundary",
|
||
"next_steps"
|
||
]
|
||
},
|
||
{
|
||
"step_id": "step_06_return_to_short_money_after_overview",
|
||
"title": "Short money follow-up after overview keeps 2020 context",
|
||
"question": "а если коротко, сколько заработали деньгами без топов?",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"2020",
|
||
"47[\\s.]*628[\\s.]*853",
|
||
"43[\\s.]*763[\\s.]*351",
|
||
"3[\\s.]*865[\\s.]*501"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive",
|
||
"285[\\s.]*819[\\s.]*547",
|
||
"147[\\s.]*855[\\s.]*827",
|
||
"Комитет государственных услуг",
|
||
"Что проверить дальше"
|
||
],
|
||
"criticality": "critical",
|
||
"semantic_tags": [
|
||
"compact_after_overview",
|
||
"temporal_carryover",
|
||
"no_all_time_leak"
|
||
]
|
||
},
|
||
{
|
||
"step_id": "step_07_plain_money_in_out_net",
|
||
"title": "Plain money request suppresses broad overview",
|
||
"question": "не обзор, просто деньги: пришло, ушло, нетто за 2020",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"47[\\s.]*628[\\s.]*853",
|
||
"43[\\s.]*763[\\s.]*351",
|
||
"3[\\s.]*865[\\s.]*501"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive",
|
||
"Комитет государственных услуг",
|
||
"12[\\s.]*093[\\s.]*465",
|
||
"12[\\s.]*792[\\s.]*194",
|
||
"Что проверить дальше"
|
||
],
|
||
"criticality": "critical",
|
||
"semantic_tags": [
|
||
"direct_money_only",
|
||
"ranking_suppression"
|
||
]
|
||
},
|
||
{
|
||
"step_id": "step_08_direct_clean_profit_2020",
|
||
"title": "Explicit clean profit goes to accounting result, not cashflow",
|
||
"question": "какая чистая прибыль по Альтернативе за 2020?",
|
||
"allowed_reply_types": [
|
||
"partial_coverage",
|
||
"factual_with_explanation",
|
||
"factual"
|
||
],
|
||
"required_answer_patterns_all": [
|
||
"7\\s*136\\s*815|7136815",
|
||
"90|91|99"
|
||
],
|
||
"forbidden_answer_patterns": [
|
||
"Учтено строк",
|
||
"Первая найденная дата",
|
||
"runtime_",
|
||
"planner_",
|
||
"query_movements",
|
||
"primitive",
|
||
"получили\\s+47[\\s.]*628[\\s.]*853",
|
||
"денежное\\s+нетто\\s+3[\\s.]*865[\\s.]*501"
|
||
],
|
||
"criticality": "critical",
|
||
"semantic_tags": [
|
||
"clean_profit",
|
||
"profit_route",
|
||
"no_cashflow_substitution"
|
||
]
|
||
}
|
||
],
|
||
"acceptance": {
|
||
"min_score": 80,
|
||
"max_unresolved_p0": 0,
|
||
"require_all_critical_steps_pass": true
|
||
}
|
||
} |