АРЧ АП11 - Вынести coverage evidence contract для exact lane и завершить phase 4 агентным прогоном

2026-04-17 11:14:10 +03:00
parent 1484c2375f
commit b3572d9d11
16 changed files with 1357 additions and 375 deletions
@@ -0,0 +1,114 @@
+{
+  "schema_version": "domain_truth_harness_spec_v1",
+  "scenario_id": "address_truth_harness_phase4_coverage_evidence_mix",
+  "domain": "address_phase4_coverage_evidence_mix",
+  "title": "Phase 4 coverage/evidence replay for counterparty fallback, inventory reset, and selected-object provenance",
+  "description": "Targeted mixed-domain replay for the explicit coverage/evidence contract: matched factual rows, blocked explanatory fallback, root reset, and temporal-limited selected-object provenance.",
+  "bindings": {},
+  "steps": [
+    {
+      "step_id": "step_01_counterparty_documents",
+      "title": "Counterparty documents use the normalized legal name",
+      "question": "покажи все документы по чепурнову",
+      "allowed_reply_types": [
+        "factual"
+      ],
+      "expected_intents": [
+        "list_documents_by_counterparty"
+      ],
+      "required_direct_answer_patterns_any": [
+        "(?i)контрагент:",
+        "(?i)чепурнов"
+      ]
+    },
+    {
+      "step_id": "step_02_counterparty_shipments_or_fallback",
+      "title": "Supplier shipment question stays human even when exact supply rows are absent",
+      "question": "что нам отгружал чепурнов, какой товар или услугу?",
+      "allowed_reply_types": [
+        "factual",
+        "partial_coverage"
+      ],
+      "expected_intents": [
+        "list_documents_by_counterparty"
+      ],
+      "required_direct_answer_patterns_any": [
+        "(?i)чепурнов",
+        "(?i)постав",
+        "(?i)оплат|возврат|товар|услуг"
+      ],
+      "forbidden_direct_answer_patterns": [
+        "(?i)^сейчас не дам прямой адресный ответ",
+        "(?i)^в текущем адресном контуре этот запрос лучше не закрывать в лоб"
+      ]
+    },
+    {
+      "step_id": "step_03_inventory_reset_march_2021",
+      "title": "Inventory root resets cleanly after the counterparty branch",
+      "question": "какие остатки на складе на март 2021",
+      "allowed_reply_types": [
+        "factual"
+      ],
+      "expected_intents": [
+        "inventory_on_hand_as_of_date"
+      ],
+      "required_filters": {
+        "as_of_date": "2021-03-31",
+        "period_from": "2021-03-01",
+        "period_to": "2021-03-31"
+      },
+      "required_direct_answer_patterns_any": [
+        "31\\.03\\.2021",
+        "(?i)на складе"
+      ],
+      "forbidden_direct_answer_patterns": [
+        "(?i)чепурнов",
+        "(?i)контрагент:"
+      ]
+    },
+    {
+      "step_id": "step_04_selected_item_supplier_temporal_limit",
+      "title": "Selected-object supplier provenance remains direct-answer-first under temporal limit",
+      "question": "По выбранному объекту \"Столешница 600*3050*26 альмандин\": кто нам это поставил?",
+      "allowed_reply_types": [
+        "factual"
+      ],
+      "expected_intents": [
+        "inventory_purchase_provenance_for_item"
+      ],
+      "required_direct_answer_patterns_any": [
+        "(?i)столешница 600\\*3050\\*26 альмандин",
+        "(?i)поставщик|поставил|куплен",
+        "(?i)союз|торговый дом"
+      ],
+      "forbidden_direct_answer_patterns": [
+        "(?i)^на 31\\.03\\.2021 на складе",
+        "(?i)^сейчас не дам прямой адресный ответ"
+      ]
+    },
+    {
+      "step_id": "step_05_inventory_same_date_restore",
+      "title": "Same-date restore returns to the March 2021 root snapshot",
+      "question": "покажи еще раз остатки на эту же дату",
+      "allowed_reply_types": [
+        "factual"
+      ],
+      "expected_intents": [
+        "inventory_on_hand_as_of_date"
+      ],
+      "required_filters": {
+        "as_of_date": "2021-03-31",
+        "period_from": "2021-03-01",
+        "period_to": "2021-03-31"
+      },
+      "required_direct_answer_patterns_any": [
+        "31\\.03\\.2021",
+        "(?i)на складе"
+      ],
+      "forbidden_direct_answer_patterns": [
+        "(?i)^сейчас не дам прямой адресный ответ",
+        "(?i)transition_not_supported_by_capability"
+      ]
+    }
+  ]
+}