Этап 4 / Волна 10: корректировка settlement-кейса — доменная фиксация синтеза, честное покрытие, удержание фокуса / Этап 4 / Волна 11: бизнес-якоря, доменное заземление и устранение утечки дебага
This commit is contained in:
Binary file not shown.
Binary file not shown.
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-001",
|
||||
"raw_question": "По каким поставщикам на конец июня не бьются взаиморасчеты, покажи документы, оплаты и хвосты.",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-002",
|
||||
"raw_question": "Сделай рейтинг самых рисковых хвостов перед закрытием периода за июнь.",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-003",
|
||||
"raw_question": "Покажи документ по номеру 000123 и строку проводки, нужен точный source-of-record.",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-004",
|
||||
"raw_question": "По 97 счету проверь, где возможна ошибка дат начала и окончания списания.",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-005",
|
||||
"raw_question": "Есть ли аномальные материалы на счете 10, которые зависли и выглядят нелогично?",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-006",
|
||||
"raw_question": "По каким реализациям 90/62 хвосты не закрылись оплатой, разложи по цепочке документов.",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-007",
|
||||
"raw_question": "Что у нас выглядит самым проблемным перед закрытием июня, если смотреть на компанию в целом?",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-008",
|
||||
"raw_question": "Покажи по банку документ №TRX-88 и связанную проводку по 51.",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-009",
|
||||
"raw_question": "Где у нас пахнет ручной ошибкой по июню?",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-010",
|
||||
"raw_question": "Какое сальдо по 68.02 за июнь?",
|
||||
"expected": {
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"case_id": "NQ-1774273974528",
|
||||
"raw_question": "По каким поставщикам не бьются взаиморасчеты по 60 счету, и разложи это по документам, оплатам и закрывающим?",
|
||||
"expected": {
|
||||
@@ -13,4 +13,4 @@
|
||||
],
|
||||
"expected_output_shape": "reconciliation_report"
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval--DLjm5dCSP",
|
||||
"timestamp": "2026-03-26T14:59:09.726Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -133,4 +133,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval--JtoGUT4Aw",
|
||||
"timestamp": "2026-03-24T10:06:13.697Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval--lulgUEKkp",
|
||||
"timestamp": "2026-03-26T15:04:48.691Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-0M5PNZp9FY",
|
||||
"timestamp": "2026-03-26T12:41:47.018Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-0sMxWQCA77",
|
||||
"timestamp": "2026-03-26T11:01:27.248Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-1I2yuUiev6",
|
||||
"timestamp": "2026-03-27T22:25:33.046Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "oKL_tEdXJzxMJ5",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "rwbRlpRZD5Ct7i",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-1ZO6LtR6Re",
|
||||
"timestamp": "2026-03-25T19:37:38.656Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-2ui-ojyBuu",
|
||||
"timestamp": "2026-03-27T23:00:27.280Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "OqFf4QS6Y8IeVm",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "qPu-4JESf8r8GX",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "GEf3pwXKJWXPD9",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-3sC2svvJlI",
|
||||
"timestamp": "2026-03-23T14:33:51.265Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -1066,4 +1066,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-3t1L3QY0wE",
|
||||
"timestamp": "2026-03-26T14:29:33.451Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-43n6WEu5ka",
|
||||
"timestamp": "2026-03-27T20:07:04.463Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "aKVQOW7oZXDoh-",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "d31RtLJyc_1TF8",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "l_EscYF5jFhLo4",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-4Jd4YjGjIL",
|
||||
"timestamp": "2026-03-26T12:55:04.731Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-4_ULiatTGc",
|
||||
"timestamp": "2026-03-27T19:54:03.827Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "fg64eigCctlx0H",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "AzrqlzdcTcOlKI",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-4xS4Pyr7Uv",
|
||||
"timestamp": "2026-03-25T20:06:37.364Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-68KSDPHUoo",
|
||||
"timestamp": "2026-03-25T19:04:30.554Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-6O6YT6OLbm",
|
||||
"timestamp": "2026-03-27T22:26:40.837Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "YG1XhXLLNtyRXi",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "UsnLAYPdib2TtP",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "OSe2GzHkTiUk_s",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-73H4_ECGyy",
|
||||
"timestamp": "2026-03-24T10:05:43.126Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-7I74UBMAdj",
|
||||
"timestamp": "2026-03-27T19:06:07.206Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "FJx9pQRvK9stcY",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Har9wm_uOvFuMw",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-7MQmPgo3_O",
|
||||
"timestamp": "2026-03-23T18:38:44.797Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -100,4 +100,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-7awjIpz1KA",
|
||||
"timestamp": "2026-03-26T15:06:11.446Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -133,4 +133,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-86xJU1J7RH",
|
||||
"timestamp": "2026-03-26T12:55:30.782Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-8HiKD7tzkR",
|
||||
"timestamp": "2026-03-26T14:55:02.675Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-94ypeytPFD",
|
||||
"timestamp": "2026-03-26T14:48:14.207Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-9KdXeQSZHM",
|
||||
"timestamp": "2026-03-27T23:01:17.627Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "xYBh3-FuKdqhv3",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "_PO5dLznW5HXGo",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-9V1JK90r5M",
|
||||
"timestamp": "2026-03-26T12:55:31.352Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-AKJhoo1T06",
|
||||
"timestamp": "2026-03-26T14:50:28.656Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -133,4 +133,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-Aowg06omhV",
|
||||
"timestamp": "2026-03-27T20:07:05.553Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Qs1sJg5iZqKR81",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "1kXfnHpTu5oDAD",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-BFy-AR9_43",
|
||||
"timestamp": "2026-03-27T19:04:16.687Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "0DaOrz69cjbPMw",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "uwOb0ginzFQerv",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "7FqVh1JZI9jRQT",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-BvJGtTd2z0",
|
||||
"timestamp": "2026-03-26T10:54:49.342Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-CBHzTg4HQQ",
|
||||
"timestamp": "2026-03-27T20:01:42.680Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "BwwaJikoWUWfnA",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "wh7Dqqtr4WcFpf",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-CcP-swdiJD",
|
||||
"timestamp": "2026-03-26T12:40:31.062Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-CsaZeqddTt",
|
||||
"timestamp": "2026-03-25T18:04:12.536Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-DBE79isARo",
|
||||
"timestamp": "2026-03-26T13:16:13.683Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-DZfBsZawGS",
|
||||
"timestamp": "2026-03-25T20:09:33.541Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-DuRnhPquJ1",
|
||||
"timestamp": "2026-03-23T18:40:07.215Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -100,4 +100,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-EDvNkoALBS",
|
||||
"timestamp": "2026-03-26T12:41:09.015Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-EJ5CZqfWe3",
|
||||
"timestamp": "2026-03-25T18:05:43.704Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-ENsXL6tmGP",
|
||||
"timestamp": "2026-03-26T14:54:03.188Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-FRq1sdepEh",
|
||||
"timestamp": "2026-03-26T12:41:47.585Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-GLNMT_DXlm",
|
||||
"timestamp": "2026-03-24T09:27:08.489Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-Geocc7VFCW",
|
||||
"timestamp": "2026-03-27T19:04:18.003Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "UmhDpl1Anoi5RC",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "xmKE3i05pSeyPF",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-HCmfC6rFKG",
|
||||
"timestamp": "2026-03-27T19:06:05.879Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "apeKyst-byAs5i",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "c2InPOuJM-3yvF",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "qzxDOPdsriwtD_",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-HK1iqqUDF3",
|
||||
"timestamp": "2026-03-26T14:55:02.664Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-HQ5HtxrQpc",
|
||||
"timestamp": "2026-03-23T18:38:22.095Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -100,4 +100,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-HrW-O78gHV",
|
||||
"timestamp": "2026-03-27T20:01:42.926Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "bs2O2VbED-XS47",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "RJi96Z0GLGl7ci",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-I9rx4KE5gR",
|
||||
"timestamp": "2026-03-26T15:06:12.667Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-II68Vzxewr",
|
||||
"timestamp": "2026-03-27T19:54:03.597Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "kJFfm4ckM0lDAh",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "b8hK0pbDYEVexZ",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-ILx0ihm-Lw",
|
||||
"timestamp": "2026-03-26T14:26:14.249Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -133,4 +133,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-IpucvMroko",
|
||||
"timestamp": "2026-03-23T17:59:01.977Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -80,4 +80,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-Jj5FenxK1M",
|
||||
"timestamp": "2026-03-25T20:49:22.808Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-KXOFxdF-Y4",
|
||||
"timestamp": "2026-03-27T22:59:30.756Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "tCP6PiE6izBcnB",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "txOhTO5Gqj8uSh",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "XBjVMKGSW93bqz",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-KoIThx--qy",
|
||||
"timestamp": "2026-03-25T20:09:40.107Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-LYEATvWSQn",
|
||||
"timestamp": "2026-03-26T14:37:56.811Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-LyUcB42oK9",
|
||||
"timestamp": "2026-03-23T15:21:58.788Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -276,4 +276,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-M5jMc2B9tS",
|
||||
"timestamp": "2026-03-24T08:24:25.436Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-MBOR7JUUoD",
|
||||
"timestamp": "2026-03-26T11:47:35.497Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-N6sUcBMVLe",
|
||||
"timestamp": "2026-03-26T10:54:58.575Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-NY6Hv8Xe_O",
|
||||
"timestamp": "2026-03-27T19:01:38.269Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "63DcJ8LUqLVrFm",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "9vQJM5Vy6-klQZ",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-NdHWJrLVPY",
|
||||
"timestamp": "2026-03-26T12:55:31.352Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-Np5p0yPVr4",
|
||||
"timestamp": "2026-03-24T11:00:15.280Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-O360CBYk6a",
|
||||
"timestamp": "2026-03-26T14:59:11.293Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-OP8Go8cQU9",
|
||||
"timestamp": "2026-03-26T13:14:25.915Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-PBFcuAprq9",
|
||||
"timestamp": "2026-03-25T20:41:15.150Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-PKpQU8_Mgt",
|
||||
"timestamp": "2026-03-26T14:48:14.219Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-Pm5Ja_gY1X",
|
||||
"timestamp": "2026-03-26T12:55:35.823Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-Pwz7xRnbx0",
|
||||
"timestamp": "2026-03-26T12:55:35.822Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-QFGmXAHVxo",
|
||||
"timestamp": "2026-03-26T13:16:30.497Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-QTRrUNA92U",
|
||||
"timestamp": "2026-03-26T13:16:30.501Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-QVtArhZLse",
|
||||
"timestamp": "2026-03-27T22:25:31.560Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Z0ep6wozOzIksr",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "GQlITuemE_XqSy",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "lyPgtBo-UKx7gz",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-QnXjMWuNG4",
|
||||
"timestamp": "2026-03-26T14:54:02.477Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -133,4 +133,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-Rgb_3vYT7U",
|
||||
"timestamp": "2026-03-23T15:23:22.517Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -230,4 +230,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-ShSTDlSKMR",
|
||||
"timestamp": "2026-03-25T20:41:15.680Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-T3ajSgm6vp",
|
||||
"timestamp": "2026-03-27T22:26:42.080Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "PT7S4E_uN0RzLx",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "2FV3EcNnTqpQ5E",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-T5ubfTDRvc",
|
||||
"timestamp": "2026-03-26T15:04:48.689Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-UrYNuoIte3",
|
||||
"timestamp": "2026-03-26T15:04:46.343Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -133,4 +133,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-VBCi_Qxr5Y",
|
||||
"timestamp": "2026-03-26T13:16:29.901Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-Vgie6SCY1Z",
|
||||
"timestamp": "2026-03-24T10:18:46.934Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-Vh9Td_UqaM",
|
||||
"timestamp": "2026-03-26T11:47:36.816Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-WWb7wzqwE2",
|
||||
"timestamp": "2026-03-23T15:31:25.602Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -1030,4 +1030,4 @@
|
||||
"request_count_for_case": 1
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-WeSd-iJ_P_",
|
||||
"timestamp": "2026-03-26T14:55:01.958Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -133,4 +133,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-WezQxQD1Jb",
|
||||
"timestamp": "2026-03-24T11:52:42.608Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -132,4 +132,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-WhK3JjwEEY",
|
||||
"timestamp": "2026-03-27T22:56:49.589Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "O59bLjW3PLR4y2",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "7UU5XbMKqPL-L-",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-X1q04xpajb",
|
||||
"timestamp": "2026-03-26T13:14:26.541Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -108,4 +108,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-X8qqhZM11T",
|
||||
"timestamp": "2026-03-27T19:04:17.793Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "8M0cpsvmIVuyI7",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "IxWaTdPLrow6-s",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -1,4 +1,4 @@
|
||||
{
|
||||
{
|
||||
"run_id": "eval-X_CQK9BieE",
|
||||
"timestamp": "2026-03-23T18:37:36.909Z",
|
||||
"mode": "single-pass-strict",
|
||||
@@ -100,4 +100,4 @@
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-XabSeNT4PM",
|
||||
"timestamp": "2026-03-27T23:00:28.773Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "yvcEYGB4rH7In4",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "duLpxzSE_VG47Y",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,137 @@
|
||||
{
|
||||
"run_id": "eval-Xb-aZzHyUg",
|
||||
"timestamp": "2026-03-27T20:01:41.552Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 66.67,
|
||||
"no_route_fragment_rate": 33.33,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 66.67,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"hybrid_store_plus_live": 1,
|
||||
"no_route": 1,
|
||||
"batch_refresh_then_store": 1
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "q1JbFvEzIhZzzi",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "d_tjl5lC47WtHQ",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "qqLW4njQ033-pr",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
Some files were not shown because too many files have changed in this diff Show More
Reference in New Issue
Block a user