Закрепить агентный replay по tUy7 smalltalk и НДС
This commit is contained in:
@@ -0,0 +1,172 @@
|
||||
{
|
||||
"schema_version": "domain_truth_harness_spec_v1",
|
||||
"scenario_id": "agent_tuy7_smalltalk_vat_purchase_repair",
|
||||
"domain": "assistant_stage1_agent_repair",
|
||||
"title": "AGENT repair replay for tUy7 smalltalk and VAT purchase-date anchor",
|
||||
"description": "Targeted replay for assistant-stage1-tUy7-NlyI5 defects: first-turn smalltalk must stay deterministic without backend_error, and VAT follow-up over the selected inventory item must keep the purchase-date anchor instead of answering with a generic organization tax period.",
|
||||
"bindings": {},
|
||||
"steps": [
|
||||
{
|
||||
"step_id": "s01_smalltalk",
|
||||
"title": "First turn smalltalk does not hit backend error",
|
||||
"question": "приветик - че как там дела",
|
||||
"required_answer_patterns_any": [
|
||||
"(?i)привет|дела|помоч|готов"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)внутренн.*ошиб",
|
||||
"(?i)контуров LLM",
|
||||
"(?i)OpenAI API key",
|
||||
"(?i)backend_error"
|
||||
],
|
||||
"criticality": "critical",
|
||||
"semantic_tags": [
|
||||
"smalltalk_entry",
|
||||
"backend_error_guard"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s02_capability",
|
||||
"title": "Capability meta remains useful after smalltalk",
|
||||
"question": "расскажи что можешь интересного",
|
||||
"required_answer_patterns_all": [
|
||||
"(?i)могу|помог",
|
||||
"(?i)ндс|склад|контрагент|долг|документ"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)backend_error",
|
||||
"(?i)capability_id",
|
||||
"(?i)runtime_"
|
||||
],
|
||||
"criticality": "important",
|
||||
"semantic_tags": [
|
||||
"capability_meta"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s03_select_org",
|
||||
"title": "Organization selection keeps working scope",
|
||||
"question": "АЛЬТЕРНАТИВА",
|
||||
"required_answer_patterns_any": [
|
||||
"(?i)Альтернатива Плюс",
|
||||
"(?i)фиксир|рабоч.*организац"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)backend_error"
|
||||
],
|
||||
"criticality": "important",
|
||||
"semantic_tags": [
|
||||
"organization_selection"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s04_stock_current",
|
||||
"title": "Inventory root establishes stock context after organization selection",
|
||||
"question": "кайф - что там на складе по остаткам?",
|
||||
"required_answer_patterns_all": [
|
||||
"(?i)склад",
|
||||
"(?i)остат|позици"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)backend_error",
|
||||
"(?i)capability_id"
|
||||
],
|
||||
"criticality": "important",
|
||||
"semantic_tags": [
|
||||
"inventory",
|
||||
"context_root"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s05_history_capability",
|
||||
"title": "Historical inventory capability follow-up",
|
||||
"question": "а исторические остатки на другие даты умеешь?",
|
||||
"required_answer_patterns_all": [
|
||||
"(?i)историческ|дат|период",
|
||||
"(?i)остат"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)backend_error"
|
||||
],
|
||||
"criticality": "important",
|
||||
"semantic_tags": [
|
||||
"inventory_history_capability"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s06_july_2017",
|
||||
"title": "Historical inventory July 2017",
|
||||
"question": "давай на июль 2017",
|
||||
"required_answer_patterns_all": [
|
||||
"(?i)31\\.07\\.2017|июл",
|
||||
"(?i)Рабочая станция"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)backend_error"
|
||||
],
|
||||
"criticality": "important",
|
||||
"semantic_tags": [
|
||||
"inventory_history"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s07_march_2016",
|
||||
"title": "Historical inventory March 2016",
|
||||
"question": "март 2016",
|
||||
"required_answer_patterns_all": [
|
||||
"(?i)31\\.03\\.2016|март",
|
||||
"(?i)Рабочая станция"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)backend_error"
|
||||
],
|
||||
"criticality": "important",
|
||||
"semantic_tags": [
|
||||
"inventory_history",
|
||||
"temporal_carryover"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s08_selected_purchase_trace",
|
||||
"title": "Selected workstation purchase trace establishes purchase-date bundle",
|
||||
"question": "По выбранному объекту \"Рабочая станция универсального специалиста (индивидуальное изготовление)\": где взяли это?",
|
||||
"required_answer_patterns_all": [
|
||||
"(?i)Рабочая станция",
|
||||
"(?i)поставщик|поступлен|закуп",
|
||||
"(?i)2015|05\\.02\\.2015|22\\.07\\.2015"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)backend_error",
|
||||
"(?i)capability_id"
|
||||
],
|
||||
"criticality": "critical",
|
||||
"semantic_tags": [
|
||||
"selected_object",
|
||||
"purchase_trace"
|
||||
]
|
||||
},
|
||||
{
|
||||
"step_id": "s09_vat_purchase_date",
|
||||
"title": "VAT follow-up keeps selected item purchase-date anchor",
|
||||
"question": "ндс можешь прикинуть на дату покупки рабочей станции?",
|
||||
"required_answer_patterns_all": [
|
||||
"(?i)НДС",
|
||||
"(?i)покуп|закуп|поступлен",
|
||||
"(?i)2015|05\\.02\\.2015|22\\.07\\.2015"
|
||||
],
|
||||
"forbidden_answer_patterns": [
|
||||
"(?i)^Коротко:",
|
||||
"(?i)2\\.631\\.388,00",
|
||||
"(?i)общ.*налогов.*позици",
|
||||
"(?i)backend_error",
|
||||
"(?i)capability_id"
|
||||
],
|
||||
"criticality": "critical",
|
||||
"semantic_tags": [
|
||||
"vat",
|
||||
"purchase_date_anchor",
|
||||
"selected_object_followup"
|
||||
]
|
||||
}
|
||||
]
|
||||
}
|
||||
Reference in New Issue
Block a user